Coder Legion - Recent posts tagged education

AI Lesson Planning for Teachers: How I Cut Prep Time in Half Without Losing the Human Touch

Sat, 31 Oct 2026 06:18:44 +0000

AI Lesson Planning for Teachers: How I Cut Prep Time in Half Without Losing the Human Touch

I am a head teacher at a government primary school in Punjab, and for years my evenings looked the same: a stack of notebooks, a syllabus outline, and lesson plans written by hand long after the school day ended. When I first tried AI lesson planning for teachers, I expected a gimmick. What I got instead was the single biggest change to my working week in years — my preparation time roughly halved, while my lessons actually got better. The catch, and it matters, is that the AI does the drafting and I do the teaching. Get that order backwards and you get slick-looking plans that fall apart in front of thirty children.

This is the honest version: the workflow I use, the prompts that work, and the guardrails that keep the human in charge.

What AI Lesson Planning for Teachers Actually Looks Like in a Real Classroom

Forget the marketing version where you type one sentence and a perfect lesson appears. Real AI lesson planning for teachers looks like this: you give the AI your topic, your class level, your time limit, and what your students already know, and it hands back a structured first draft — objectives, a timed sequence, questions, and a short activity. Then you do the part no model can do: you cross out what won't work for your children, you fix the examples so they match your students' lives, and you add the one story or question that makes the class pay attention.

In my school, a typical lesson is 35 to 40 minutes with a large class and limited materials. An AI that doesn't know that will happily plan a lesson needing a projector, printed worksheets for every child, and group work in sixes. So the first skill isn't prompting. It's knowing your constraints well enough to state them: class size, time, available materials, language level, and what the textbook actually covers. The AI is only as grounded as the brief you give it.

One more honest point: I teach from the Punjab textbooks, and the AI does not know my exact chapter — only the general topic. That gap is where a careless teacher gets into trouble, accepting a plan with the wrong scope or examples. My rule: the textbook is the syllabus, the AI is the assistant. Every plan gets checked against the chapter before it reaches the classroom.

My 30-Minute AI Lesson Planning Workflow

Here is the full workflow, timed honestly. It replaced an evening routine that used to take over an hour.

Step 1 — Pick one clear objective (3 minutes). Not "teach fractions" but "by the end of the lesson, students can shade and name one-half, one-third, and one-quarter of a shape." One objective, stated in student language. If you can't write it in one sentence, the AI can't plan around it either.

Step 2 — Write the brief (5 minutes). I paste the same structure every time: topic, class, time available, what students already know, materials I actually have (chalkboard, chart paper, the textbook), and one constraint ("no printing", "mixed ability class", "Urdu-medium explanations where needed"). This brief is the whole secret. A vague brief gives you a generic lesson; a specific brief gives you a usable one.

Step 3 — Generate the first draft (2 minutes). I ask for a timed plan: a 5-minute hook, 15 minutes of teaching, 10 minutes of practice, and a 5-minute check. I also ask for three questions at different difficulty levels and one activity that needs no materials beyond the board. The draft is never the final plan. It is raw material.

Step 4 — Edit like a teacher, not an editor (15 minutes). This is where the real work happens. I replace generic examples with local ones — market prices, cricket scores, the school garden — because children remember what they recognize. I delete any activity I can't run with my class size, check every fact and worked example against the textbook, and add the opening question I know will hook this particular class, because no model knows my students.

Step 5 — Make the exit ticket (5 minutes). Every lesson ends with one quick check: two questions on the board, answered on slates or in notebooks, that tell me before the bell whether the objective landed. If I skip this step, I am guessing about learning. The AI drafts the questions; I pick the ones that actually test the objective.

Total: about 30 minutes.

The Prompt Pattern That Keeps Lessons Honest

Most bad AI lessons come from one-prompt thinking: one long prompt, one accepted answer. A three-pass pattern fixes most of the quality problems people blame on the AI.

Pass 1 — Draft. "Plan a 35-minute lesson on [topic] for class [X]. Students already know [Y]. Materials: chalkboard and textbook only. Give me timed sections, three questions at easy, medium, and hard levels, and one no-material activity."

Pass 2 — Localize. "Rewrite the examples using everyday life in a Pakistani village school: market shopping, farming, cricket, family routines. Replace any activity that needs printing or a projector."

Pass 3 — Challenge. "Now find the weak points in this plan: where will students get confused, which step is too fast for a mixed-ability class, and what question checks understanding after the main teaching?" That third pass is the most valuable one — it surfaces the exact spots where a real class will struggle, and trains you to spot them yourself over time.

One warning from experience: never let the AI write your assessment questions unchecked. Models produce confident-sounding questions with wrong answers, especially in maths and science. Every question gets solved by me before it reaches the students. That single habit is the difference between using AI and being used by it.

AI Lesson Planning for Teachers: The Guardrails I Never Skip

AI lesson planning for teachers only stays honest if you hold a few lines firm. These are mine, learned the hard way:

The textbook decides the scope. If the AI's plan covers something outside the chapter, it gets cut, however interesting it looks. Exam alignment and student confidence both depend on this.
Every fact gets verified. Dates, science facts, worked maths examples, spellings — all checked before class. I once caught a wrong fraction example in a draft five minutes before the lesson. That was the day this rule became non-negotiable.
No student data goes into the AI. No names, no marks, no photos, no personal details. I describe the class in general terms ("mixed ability, 30+ students") and nothing more. This is a privacy line, not a preference.
The AI never writes the final word to students. Worksheets, notes, and homework pass through my eyes first. Anything the children read should sound like their teacher, not like a chatbot.
One new thing per lesson. If the AI suggests a new method and a new activity and a new grouping in the same lesson, I keep one and drop the rest. Classrooms punish overload faster than any reviewer will.

These guardrails are why my prep time halved instead of disappearing. The AI removed the blank-page work — structuring, sequencing, drafting questions — and left the judgment work to me. That is why the lessons got better, not just faster.

A Real Example: One Lesson, Before and After

Take a class 4 maths lesson on fractions, a topic where children either click or drown. My old hand-written plan was: introduce the term, draw two shapes on the board, do three textbook exercises. Serviceable, forgettable.

The AI-assisted version kept the same objective but changed the shape of the lesson. The hook became a real question: "If four friends share two rotis equally, how much does each get?" — something every child in my class has actually argued about. The teaching section used folding chart paper instead of just drawing, so children could see and hold the halves and quarters. Practice became a quick market game: "A shopkeeper sells half a kilo of sugar, then a quarter — how much is left?" And the exit ticket was two questions, one easy and one that catches the classic mistake of confusing the number of pieces with the size of each piece.

Nothing in that lesson required classroom technology. The AI contributed structure and variety; the classroom contributed the rotis, the market, and the paper-folding. AI lesson planning for teachers works best when the technology stays in the preparation room and the classroom stays human.

Your First-Week Checklist

If you want to try this yourself, here is the honest starter plan — one week, no overwhelm:

Pick ONE subject and ONE class to experiment with. Don't AI-plan everything at once.
Write your brief template once (topic, class, time, prior knowledge, materials, constraints) and reuse it every day.
Run the three-pass pattern: draft, localize, challenge.
Check every fact and every worked example against the textbook before class.
End every lesson with a two-question exit ticket, and keep the results — they tell you within a week whether the plans are actually working.
Steal the good parts: when a hook or activity lands well, save it in a personal bank for next year. After a term, that bank is worth more than any prompt collection.

A practical note for fellow Pakistani teachers: the free Punjab textbooks are all available online now, so you can read the chapter on your phone while drafting. I keep the full library linked from the Punjab textbooks hub on rehmanahmad.com, and I grab classroom utilities like PDF tools from filecobra.com when I need a printable for class. The tools support the teaching; they never replace it.

The Bottom Line

AI lesson planning for teachers is not about letting a machine teach. It is about letting a machine do the drafting so the teacher can do the teaching — better prepared, less exhausted, with the evening back. My prep time roughly halved, my lessons got more varied, and my students got hooks I would not have invented on a tired evening with a blank page.

But the bargain only holds if you keep the guardrails: textbook as syllabus, every fact verified, no student data shared, and your judgment on every page that reaches a child. Do that, and the AI becomes what it should have been all along — a tireless assistant to a teacher who is still, unmistakably, in charge.

If you try this workflow, start with one lesson this week and judge it by the exit ticket, not by how impressive the plan looked on screen.

The Ultimate Guide to Online Education Platforms with Analytics for Coaching Institutes

Fri, 30 Oct 2026 05:33:49 +0000

The way in which coaching institutes teach manage students. Measure academic progress is changing fast. Today’s learners expect flexibility, timely feedback and access to learning resources that go beyond the classroom. At the time coaching institutes have to manage batches, attendance, performance and show meaningful learning outcomes.

This can become a challenge to handle all these responsibilities manually especially when an institute has courses, hundreds of students and growing faculty team. This is where online education platforms with analytics can really make a difference.

These platforms combine learning management and performance tracking which enables coaching institutes to make data driven decisions boost student engagement and streamline operations. With so many solutions available how does an institute make the right choice?

Knowing the features, benefits and selection criteria can help you identify the best online education platform for coaching institutes with analytics and build a more effective data-driven learning environment.

What is an online education platform with analytics?

An online education platform is a learning environment where teachers are able to deliver courses, share study materials, assess learners and interact with students. If the platform has analytic capabilities it can also display useful data on student participation, academic performance and learning progress.

Coaching institutes can move beyond attendance registers, spreadsheets or periodic test results to a centralized platform to get insights on how students are interacting with learning content and where they need additional support.

For instance an institution could observe that multiple students are struggling with the subject or that attendance declines at a particular point in a course. Learning analytics can assist educators in detecting patterns and exploring the reasons for them.

The aim is not just to gather data. Its purpose is to transform information into actionable decisions that enhance teaching and learning.

Why Learning Analytics Should Be Adopted By Coaching Institutes?

All coaching institutes want their students to do well stay engaged and finish their courses successfully. Without reliable information it is difficult to find the reasons for the inconsistent performance.

Analytics helps fill this gap shedding light on student activity and academic progress. When teachers are able to identify knowledge gap they can step in before small challenges become large learning barriers.

More Insight into Student Performance

Test scores are a measure of academic progress but they don’t always tell the story behind why a student is struggling. A student might understand the ideas. Make mistakes under pressure of time. Another may perpetually miss questions from a given topic because the underlying fundamentals are murky.

Assessment analytics in a platform help the educators to look at the performance across the subjects, topics, tests and batches. This helps to identify problems that recur and change the teaching approach accordingly.

Enhance Student Engagement

You can also use attendance, assignment completion, participation and learning activity as measures of engagement. When all of these indicators are in one place faculty can identify students who may be falling behind.

If participation drops suddenly for instance a teacher can figure out if a student is having trouble or needs more support.

Analytics can’t account for every change in behavior but it can help educators to recognize when a conversation or intervention might be appropriate.

Smarter College Choices

Coaching institutes decide batch allocation, revision schedules, faculty support, extra practice sessions regularly. Without information these decisions can be highly assumption-based or isolated observations.

Analytics provide a structured basis for assessing what’s working and where there’s scope for improvement. Over time institutes may utilize this insight to improve their processes and provide a more consistent learning experience.

Important Features to Consider in an Online Education Platform

The choice of an education platform is not simply a matter of comparing the cost of subscriptions or verifying that online classes are supported. The right solution should align with the institute’s teaching model, operational needs and long-term growth plans.

Complete Learning Management

A good platform should allow you to organize courses upload study materials, schedule classes and manage learning resources. Students should be able to access content without having to wade through complex systems.

Course and batch management is very important for coaching institutes offering programs or preparation for competitive examinations. A good learning environment keeps students on track. Reduces repetitive administrative work for faculty members.

Analytics of Performance and Assessment

Assessment tools should consist of tests, quizzes, assignments and performance tracking. Ideally the educator should be able to see the results of a student as well as trends across batches and subjects.

Useful analytics can include test scores, topic-wise accuracy, assessment completion and changes in performance over time. Specific capabilities vary by platform and institutes should assess whether the reports suit their academic needs.

The best online education platform for coaching institutes with analytics should make these insights easy to understand and act upon without teachers with complex dashboards.

Student Support and Communication

Regular communication keeps students informed of classes, assignments, test schedules and academic expectations. Institutes should seek features that allow sharing of updates and timely provision of guidance.

These capabilities may include notifications, announcements, discussion spaces or other communication tools depending on the platform. The conclusion is that the right info is readily available to students and teachers without a lot of hassles.

Data Security and Access Control

Education platforms will store student information, assessment records and sometimes payment or contact details. This information should therefore be a consideration in platform selection and protected.

Institutes should consider access rights, authentication, data protection standards, backup arrangements and the provider’s privacy position. They should also have internal procedures for determining who has access to student records and reports.

A feature-rich platform is only valuable if used with the security and privacy practices.

The Role of Analytics in Improving Teaching and Learning Outcomes

The true meaning of analytics is when teachers use data to enhance the learning process.

Suppose there is a coaching institute which conducts a mathematics test The overall batch average seems satisfactory but if we look at topic wise results many students are finding algebra difficult. The faculty team can instead go back. Revisit the concept provide additional practice and do a follow-up assessment instead of just moving on to the next chapter.

Analytics may also reveal that students who always do the practice assignments tend to perform on the mock tests. Faculty can use this observation to motivate practice and identify students who may require additional support.

Analytics can also help institutes understand how successful their teaching methods are. When students keep having problems with a topic over batches it might be worthwhile to examine the teaching methodology learning resources or the test format.

These examples show how data can lead to academic decision making. Analytics should support teachers’ professional judgments, not replace them. The context of the student of interactions in the classroom and of individual learning needs are critical to understanding performance.

Best Online Education Platforms in India for Coaching Institutes – How to Choose?

The Indian coaching industry caters to a set of learners from school students preparing for board examinations to aspirants preparing for competitive entrance and government recruitment examinations. Thus the need of one institute may be quite different from those of another institute.

While judging the best online education platforms in India for coaching institutes know what problems your institute needs to solve first.

An institute focused on classes would put emphasis on video delivery, scheduling and interactive learning tools. A test-prep center might focus more on question banks, practice tests, automated scoring and detailed performance reports. If you’re an organization with multiple branches you may require centralized administration, role-based access and consolidated reporting.

Scalability is another factor. A good platform for a batch should ideally be able to handle more students, courses and faculty without creating unnecessary operational complexity.

Ease of use is also important. A sophisticated system will be of value if it is difficult for teachers to manage or for students to navigate. Institutions should consider the student and faculty experience before committing term.

Pricing should be taken in context well. Check subscription costs, implementation requirements, support arrangements and additional charges for some functionalities. Of going for the cheapest option look for a platform that provides the best mix of features ease of use and long-term benefits.

Finally consider the level of customer support and the provider’s ability to help your team onboard to the system. A smooth implementation process can make a difference to the platform’s overall effectiveness.

Mistakes to Avoid When Selecting an Education Platform

A mistake is to pick a platform based on the number of features it has. More features don’t automatically mean results. Institutes should concentrate on the capabilities that're directly responsive, to their teaching and administrative needs.

Another mistake is to ignore how useful analytics can be. A dashboard can show a lot of information. The institute needs to be able to understand the data and use it to make good decisions. Before you pick a platform look at sample reports. Check if they answer the questions that your faculty members usually ask.

Institutes also must not use technology in place of teaching. Even though an education platform can help learning by being organized easy to access and giving insights the main thing that helps students grow is teaching, regular practice, proper feedback and real interaction between teachers and students.

Finally don't forget about how to put the system into place and train people. Teachers need time to learn the system set up their ways of working and use the features in their teaching. Letting teachers be part of choosing the platform can help them use it better and make sure the platform fits what happens in classrooms.

The Future of Coaching Centers: Data-Driven Learning

As digital education keeps changing analytics is expected to become more important in how coaching institutes understand what students need and provide help.

Newer systems could give detailed learning insights, personalized suggestions and automatic reports. Artificial intelligence could also help teachers see patterns in a lot of test data. The quality of the insights would depend on the quality of the data and how well the tools are used.

Institutes should also have a way of doing things. Students’ data needs to be treated, automated suggestions should be checked and teachers should stay involved in important academic choices.

The goal is not to make education fully based on data. It is to mix the data with human understanding so that every student gets better guidance and more chances to improve.

Transforming Learning Data into Improved Student Outcomes

For any coaching institute that wants to improve its efficiency, student involvement and make academic choices picking an online education platform with analytics is an important choice. The best solution includes learning materials, tests, tracking progress and helpful information to support both teachers and students.

Think about what your institute wants before making any decision. Compare the features ways to report how easy it is to grow, security, how easy it is to use and support. Important make sure the platform helps your teachers take information and use it to make teaching better.

ItLearn helps coaching institutes move to an online education setup to manage teaching and learning. If your institute is ready to make academic work easier improve the learning experience and find better ways to track student progress contact ItLearn now. Find out how the right education technology can help your institute build an organized, interesting and ready, for the future learning experience.

Can we learn Agentic AI by playing games?

Thu, 29 Oct 2026 05:46:33 +0000

Current educational platforms focus mainly on content

However, the problem with content is that, no matter how good the content is, it's a passive medium for learning.

A proven way to learn is to participate actively.

With that in mind, we are developing an educational game: AI Detective - a story-based game to learn Agentic AI.

We will be releasing it soon.

If you need early access, send a message in a DM or comment.

MarcX Studio

Tue, 27 Oct 2026 23:05:38 +0000

MarcX Studio is an independent studio making simple education and language apps for iPhone. None of the apps are AI products: each does one job clearly, most work without an account, and your data stays on your device unless you choose to back it up or share it.

Apps

RooCards: keep your loyalty cards ready on your iPhone.
Snap Spell: spelling practice from the word lists students bring home. Scan or paste a list and practise it.
NumListen: number listening practice for language learners.
成语拾光 (Idiom Shiguang): a Chinese idiom game.

The apps are free to download, with optional upgrades for people who want more. The studio site, https://www.marcx.app, is available in many languages, including English, Chinese, Japanese, Korean, Spanish, French and German.

Website: https://www.marcx.app

GitHub Classroom has closed: what you still have, and how to run next term

Mon, 26 Oct 2026 16:09:06 +0000

Disclosure: I work on Uppcoder, one of the options listed at the end. The rest is about GitHub Classroom itself and applies whatever you choose.

GitHub retired GitHub Classroom on 28 August 2026. If your course ran on it, here is what is still there, what is gone, and what your choices are for the next term.

What happened, and when

GitHub closed new sign-ups to GitHub Classroom in May and retired it at the end of August. It now points educators to two partners: Classroom 50, free and open source from the Fifty Foundation, the nonprofit behind CS50, and Codio, a commercial platform.

26 May 2026: new sign-ups to GitHub Classroom closed
28 August 2026: GitHub Classroom retired; its website, APIs and related services decommissioned
GitHub's recommended partners: Classroom 50 and Codio

What you still have

GitHub says that GitHub accounts, repositories and organisations created to work with Classroom are not affected. Everything that lived in a repository is still where it was.

Your classroom organisation
Every student repository Classroom created, with its full commit history
Your template (starter) repositories
Any tests kept inside a repository, for example in a tests folder or a GitHub Actions workflow

What is gone, or going

GitHub says some Classroom data is being permanently deleted under its retention policies. GitHub's Classroom Export Utility downloaded this data as JSON and CSV files, but it worked through the Classroom APIs, which GitHub says are now decommissioned. If you ran it before 28 August, keep those files; if you did not, plan to rebuild what is listed here.

Classroom and assignment names
Tests defined in Classroom itself rather than in a repository (autograding set up on the Classroom website)
LMS-integrated rosters: the LTI link between Classroom and Moodle, Canvas and the rest

Finding your students' repositories

Classroom named each student repository after the assignment and the student's GitHub username, so the repositories of one assignment share a prefix. With the GitHub CLI (gh) you can list them, and clone them all into one folder. The roster that matched usernames to students was Classroom data; rebuild it from your LMS class list or your own records if you need it.

# list one assignment's repositories
gh repo list YOUR-ORG --limit 1000 --json name -q '.[].name' | grep '^assignment-prefix-'

# clone them all into the current folder
gh repo list YOUR-ORG --limit 1000 --json name -q '.[].name' | grep '^assignment-prefix-' \
  | xargs -I{} gh repo clone YOUR-ORG/{}

Before next term: a checklist

Whatever you use next, these steps keep last term's work safe and make the next one easier to set up.

Keep the organisation. Archive last term's student repositories rather than deleting them, if your school's retention rules allow it
Move tests into the template repositories, in a tests folder, so they travel with the assignment whatever tool you use next
Write down the assignment list with each template repository and its due dates, since the names in Classroom are being deleted
Check your LMS: if Classroom was linked to Moodle, Canvas, Blackboard or Brightspace, remove the old activity so students don't click into a dead end
Pick the replacement before the term starts and try one assignment end to end with a test student account

The options for next term

Each takes a different approach; which fits depends on your course, your school's rules on student data, and how much you want to run yourself.

Classroom 50 (free, open source, the Fifty Foundation): closest to GitHub Classroom, with assignments from template repositories, autograding through GitHub Actions, rosters and submissions, on GitHub. Needs an organisation on a GitHub Team or Enterprise plan (free for verified educators through GitHub Education) and a token with read and write access to the organisation's repositories
Codio (commercial, a GitHub partner): a hosted platform for coding courses with its own content, environments and autograding, and migration guides for GitHub Classroom users
Do it yourself with the GitHub CLI: for small courses, create the student repositories from a template with gh and a list of usernames, and mark by hand or with your own GitHub Actions; community tools shared in GitHub's own discussion of the change do this for you
Uppcoder (that's us; commercial, with a free tier): assignments are git repositories with starter code, tests and a solution; every student works in the same ready environment in the cloud, so nobody installs anything; tests mark each hand-in and you mark the rest by rubric; hand-ins can stay in your own GitHub organisation; it works inside Moodle, Canvas, Blackboard and Brightspace over LTI 1.3, or on its own;

If you choose Uppcoder: for teachers moving from GitHub Classroom we move one course across for you, up to ten of its assignments, and it runs free until 30 June 2027 (first 20 courses, one per teacher, up to 40 students, apply by 15 December 2026). Details: https://portal.uppcoder.com/github-classroom-alternative?utm_source=coderlegion&utm_medium=blog&utm_campaign=posts&utm_content=post-146

Sources

https://portal.uppcoder.com/github-classroom-closed?utm_source=coderlegion&utm_medium=blog&utm_campaign=posts&utm_content=post-146

The Rs 20 Pencil Bank: How My Class 2 Stopped Losing Pencils Every Day

Mon, 26 Oct 2026 04:49:31 +0000

The Rs 20 Pencil Bank: How My Class 2 Stopped Losing Pencils Every Day

From The Head Teacher's Diary — real stories from a primary-school classroom in Yazman, Pakistan. Entry No. 1.

Every morning, the same complaints waited for me at the classroom door.

A pencil was lost. An eraser had disappeared. A sharpener was nowhere to be found. And behind those small complaints stood a bigger problem: the parents of my students are poor. A pencil costs little, but when a child loses stationery almost every day, "little" stops being little. Children were taking money from home daily for replacements — and at home, that money came out of an already tight budget.

I am the head teacher of my primary school, and also the in-charge of Class 2. So this was my problem to solve, twice over.

The Problem Was Not the Children

The easy response is to scold. "You are careless. Look after your things." Every teacher has said it; I have said it myself.

But watch the situation honestly: a six or seven year old, a full day of lessons, playtime, bags being opened and closed a dozen times — and no place, no system, no routine for keeping three small items safe. We were expecting small children to protect stationery with nothing but memory and fear of scolding.

Teachers everywhere know this as the problem of disappearing pencils — and it is rarely solved by buying more pencils.

The children were not the problem. The absence of a system was.

So instead of another lecture, I built a system.

Every Child Gives Rs 20 — Once

I told the class: every child will bring Rs 20. Not every day. Just once, into one shared collection.

From that collection, we bought pencils, erasers and sharpeners for the whole class — together with a small bin to keep them in. From that day, the stationery stopped belonging to individual children. It belonged to the class — community supplies, owned by everyone and cared for by everyone.

Then came the part that made it work: two students are on duty every day.

In the morning, the duty students hand out a pencil, an eraser and a sharpener to every child.
At closing time, they collect everything back from everyone and place it in the bin.
The next morning, the same routine begins again.

That is the whole system. No register, no software, no expense beyond the first collection.

The whole system in one picture: one red bin, and every pencil, eraser, sharpener and ruler belongs to the class — not to any one child.

What Changed

The results surprised even me.

A pencil now lasts about two weeks. Before the pencil bank, a pencil could vanish the same day it was bought. Now the same pencil keeps coming back to the bin every evening, so it survives until it is genuinely used up.

The daily money stopped. Earlier, children asked for money at home almost every day for a lost item. Now the collection happens once, and the next one is needed only after roughly two weeks. For a poor household, removing a daily small expense is not a small thing.

The complaints ended. The morning queue of lost-pencils simply disappeared, because there is nothing left to lose — everything returns to the bin.

And something less visible grew: responsibility. The duty students take their job seriously. Collecting from everyone, missing no one, keeping the bin in order — that is trust, and children rise to meet trust. The whole class cares for the stationery now, because it is ours. A personal pencil got lost. The class pencil gets protected.

Why It Works

Looking back, three things made the difference:

Ownership changed. What belongs to everyone is watched by everyone.
The routine is daily and visible. Out in the morning, back at closing. Nothing depends on a child's memory at home.
The cost is shared and tiny. Rs 20, once in a while, never burdens one family — and no child is singled out, because everyone is an equal shareholder in the bank.

Try It in Your Classroom

If your class has the same morning complaints, the recipe is simple:

Fix a small, equal contribution every child can afford — for us it was Rs 20.
Buy the stationery for the whole class from the pooled money, plus one box or bin.
Put two students on duty daily — hand out in the morning, collect everything at closing.
Rotate the duty so every child serves.
Keep the collection and spending open and simple; the class should always know where the money went.

You do not need permission, a budget line, or an NGO. You need Rs 20 per child and the decision to fix the system instead of the child.

The Lesson I Took Home

Teaching is full of problems we blame on children: lost things, forgotten things, unfinished things. This small experiment taught me to ask a different question first — is there a system, or am I expecting a seven year old to be one?

A pencil bank will not transform education. But it turned a daily complaint into a daily lesson in sharing, duty and care — and it costs less than the price of the pencils we were losing.

If you try it in your school, I would genuinely like to hear how it went.

— Rahman Ahmad Chaudhary, Head Teacher, GPS 117 DB(BJ), Yazman. More entries from The Head Teacher's Diary are on the way.

FAQs

How do you solve the problem of disappearing pencils in the classroom?
Stop treating each pencil as private property. Pool a tiny equal contribution (we used Rs 20 per child), buy stationery for the whole class, store it in one bin, and have two duty students hand it out every morning and collect it back at closing. When stationery belongs to the class and returns to one place daily, losses almost disappear.

How do you manage classroom stationery when parents cannot afford supplies?
Spread the cost thin and share it. Our pencil bank runs on Rs 20 per child, collected roughly once every two weeks — far cheaper for poor families than replacing a lost pencil, eraser or sharpener every other day, and no child is singled out because everyone contributes equally.

What is a class pencil bank?
A pencil bank is a simple classroom system: stationery is bought from a small pooled contribution, owned by the class, kept in a shared bin, and distributed and collected daily by students on duty. In my Class 2 it stretched one pencil's life to about two weeks and ended daily lost-stationery complaints.

Every Punjab Textbook, Online and Free: How I Put 251 Books in Every Student's Pocket

Mon, 19 Oct 2026 09:14:22 +0000

Every Punjab Textbook, Online and Free: How I Put 251 Books in Every Student's Pocket

This week I finished something I'd been dreaming about for years: the entire Punjab textbooks online library is now live on my website — 251 books, every class from pre-primary to 12th grade, free for every student, teacher, and parent in Pakistan to read, download, and learn from.

I'm a head teacher at a government primary school in Yazman, Punjab. For eight years I've watched kids walk to school without a single textbook in their bag — not because they're careless, but because books get lost, torn, shared between three siblings, or were never issued at all. But their parents almost always have a smartphone. So I asked myself a simple question: if the phone is already in the house, why isn't the textbook?

The answer is now my free Punjab textbooks library. Here's the full story — why raw PDFs were never enough, what I built instead, and what I learned digitizing 251 books as a teacher who codes at night.

Why Reading Punjab Textbooks Online Needed a Real Library

The official PDFs already existed. The Punjab Curriculum and Textbook Board's E&SE page (PECTAA) hosts them on Google Drive. So why did I spend weeks building a library around them?

Because "the PDF exists somewhere" and "a student can actually read it" are two very different things. Here's what raw textbook PDFs look like in practice:

No navigation. A 200-page scanned PDF has no table of contents. Finding Chapter 3 of Science means scrolling past 60 pages of blurry scans on a 6-inch screen.
Brutal file sizes. Some of these PDFs are 100MB+. On mobile data in a village, that download dies halfway — twice.
No discoverability. The board's Drive folders are organized for bureaucrats, not for a Class 7 student looking for "Waqfiyat-e-Aama."
No learning layer. A PDF is a photocopy. Nobody asks questions about a photocopy, nobody tests you on it, nobody tells you the tricky part of Chapter 4.

I didn't want a download dump. I wanted a library — the kind of page I'd send a parent on WhatsApp and know they'd figure it out in ten seconds. So that's what I built: the Punjab textbooks library at rehmanahmad.com/textbooks.

Building the Punjab Textbooks Online Library

Thirteen grade levels. 251 books. Every single book page carries more than a download button — it carries everything a student needs to actually study the book:

Real chapter tables of contents. I typed out the chapter lists for every book from the official PDFs — including scanned ones where I had to read the fehrist (table of contents) page visually. Open the Class 8 History book and you see all 5 chapters listed, clickable, in seconds.

The "Online Parho" reader. Instead of forcing a 150MB download, each book opens in a continuous-scroll reader right in the browser: lazy page rendering (a 200-page book loads like a webpage, not like a freight train), zoom for small text, go-to-page, and fullscreen mode that turns any classroom projector into a smart board.

A chapter quiz game on every book. Five multiple-choice questions, instant score, one-tap restart. I built it because my students beg to try again when they miss two — and "begging to restudy" is the closest thing to magic a teacher gets.

Teacher tips and FAQs. Every page has practical notes written from real classroom experience — which chapters trip students up, how to pace them — plus FAQs with proper search-engine markup so parents googling "Class 9 Biology chapters" actually land on the right page.

Here's the full rollout, all verified live this week:

Grade	Books
Pre-primary	4
Class 1–2	7 each
Class 3–5	7, 8, 8
Class 6–8	19, 21, 20
Class 9–10	55, 47
Class 11–12	20, 28

Class 9 alone is 55 books — core subjects plus the technical stream. For 11th and 12th, the official PECTAA page doesn't list books at all, so I sourced them from official PCTB mirrors and ran every file through the same verification pipeline: official list match, real chapter extraction, cover art, build, and a live check before it counted as done.

How to Read Punjab Textbooks Online: A 3-Step Walkthrough

Using it is deliberately boring — three taps and you're reading:

Pick your class. Open rehmanahmad.com/textbooks and tap your class. Pre-primary through 12th, all on one hub page.
Pick your book. Every class page shows each book with its cover, subject, and medium (English/Urdu). Tap one — say Class 10 Biology — and you get the full chapter list, teacher tips, and FAQs.
Read, quiz, or download. Read in the Online Parho reader (works on any phone browser, no app needed), play the chapter quiz after the lesson, or grab the official PDF download for offline study.

That's the whole pitch. No login, no paywall, no app install — because in my school, "create an account first" is where usage goes to die.

5 Lessons From Digitizing 251 Textbooks

1. The table of contents matters more than the PDF. Nobody on mobile data downloads a 150MB file to find Chapter 3. The single most-used feature of the whole library is the chapter list. Navigation is the product.

2. Mobile-first isn't a slogan — it's the only device that exists. My students don't have laptops. Every design decision assumed a cracked-screen Android phone on 4G: lazy rendering, compressed cover thumbnails, no heavy JavaScript frameworks on the reader.

3. Quizzes turn reading into a game. I added the 5-question quiz as a nice extra. It became the headline feature. Kids who would never "revise Chapter 4" will happily play "the Chapter 4 game" four times to beat their score. Gamification works when it's five minutes and zero setup.

4. Use official sources only — and say so. Every book is matched against the official board lists, with attribution. When the official page had gaps (Classes 11–12), I used official PCTB mirrors rather than random uploads, and I say that openly on the site. In education, trust is the whole currency.

5. No thin pages, ever. My rule for everything I publish: a page with a download button and nothing else helps nobody. Every book page carries real, original content — chapter lists, teacher tips, FAQs. A library of 251 stubs would have been faster to build and worthless to use.

The 5-Minute Classroom Checklist

If you're a teacher, here's exactly how to use the library in class this week:

[ ] Open your class page on the classroom tablet or projector-connected phone
[ ] Open the book's Online Parho reader in fullscreen — instant smart board
[ ] Use zoom on diagrams and maps instead of squinting at a printed copy
[ ] After the lesson, run the chapter quiz on the big screen — let the class vote on answers
[ ] Share the book's link in the parents' WhatsApp group so kids can re-read at home

And a side note for the computer-lab side of school life: for the Windows tools my lab runs on — PDF readers, document tools, classroom utilities — I grab verified free software from filecobra.com, the free-software download directory I built and run with hash-and-signature verification on every file. Same philosophy as the library: free, safe, no junk.

What's Next: Sindh, Then the Rest of the Country

Punjab was Phase 1. Phase 2 is Sindh — the STBB board's own ebook portal already hosts 194+ books from ECCE through Class 12 in Sindhi, Urdu, and English, so the groundwork exists. Balochistan comes after that. The goal is simple and, honestly, a little stubborn: every textbook in Pakistan, online, free, readable on the cheapest phone in the village.

If you're a teacher, a parent, or a student: bookmark the library and share it with one person who needs it. That's how this grows.

And a question for you: which board should I digitize next — and what's the one feature that would make your students actually open their books? I read every comment. Teacher by day, developer by night — this is what the nights are for.

How Online Learning Can Help Coaching Institutes Go Digital

Fri, 16 Oct 2026 10:04:19 +0000

Today’s students learn very differently than students did even just a few years ago. Although many still prefer sitting in a classroom and learning directly from a teacher, others want the flexibility to study from home, revisit recorded lessons or attend classes without traveling long distances.

Coaching institutes see this change as a challenge and an opportunity. Students have never had so much choice and institutes need to find ways to meet their expectations, without compromising on the quality of teaching.

The transition to digital can help. It enables the coaching centers to reach out to students beyond the immediate neighborhood, offer courses in different formats and make learning resources available when students need them. However, going online is not only about having classes on video call. Institutes also need an easy way to organize courses, share study materials, assess students and keep everything running smoothly.

The right learning platform can make taking your institute online a practical next step rather than a daunting task.

Why Digital Transformation of Coaching Institutes?

Anyone who has run a coaching institute knows there are certain limitations of classroom teaching. The size of an institute is determined by the space, teachers and schedule. It is limited by the number of students it can accommodate. Traveling long distances each day can be hard for students who live far away, and students who miss a class might find it hard to catch up.

Digital learning addresses some of these challenges.

Students can attend online classes, study materials and recorded lectures whenever they want to revise a topic. This is especially useful for competitive exam preparation, where students often need to practice again and again, revisit concepts they find difficult.

The biggest benefit for institutes is the chance to reach more learners without having to set up another physical center. For example, a coaching institute in one city can begin online courses for students in other parts of India.

It also allows for trying out different formats of teaching. Institutes can continue with regular batches in the classroom and also start recorded courses, online revision sessions or extra practice classes.

The idea isn’t to replace classroom teaching. To provide students with more ways to learn and help institutes make better use of their teaching resources.

What Is a Good Online Learning Platform for Institutes?

Online classes are pretty easy to begin with. The complications can come in when the number of students increases, and you’re managing the entire learning experience.

Study materials can be scattered across different apps, assignments can get lost in messages, and teachers can spend too much time on administrative tasks. Students, however, may have a hard time locating the correct lesson or recalling what they need to complete.

A good online learning platform can collect these activities together.

It must be user-friendly. Students should be able to access their courses and learning materials without needing to call for technical assistance all the time. Teachers should be able to create lessons and conduct their classes without having to deal with a system that is overly complex.

Another important aspect is course management. Institutes need an easy way to organize lessons, upload resources and make learning content available to students. Assessments, quizzes and assignments can help to further support the learning process by helping students to practice and giving teachers a better understanding of their progress.

It is also worth considering what the institute will need in the future. A platform that works for a small batch should optimally be able to handle growth as more students enrolll and new courses are added.

When selecting the best online education platform for institutes to go digital, institutes should consider what actually matters to their teachers and students. The best solution is not always the one with the most features. It is the one that helps you to teach, makes learning more accessible and simplifies everyday administration.

How to Create Your Own Branded Online Academy

Imagine students who are aware of your institute’s name, have faith in your faculty, and are eager to learn from you even if they can’t attend your physical classes. An online academy provides you a way to make that happen.

Institutes can create their own digital learning environment rather than using third-party learning platforms. The institute forms its identity outside of the classroom and students can take courses associated with the institute.

For instance, a coaching center for competitive exams could provide recorded foundation courses to students from all over India. An institute of language can start online classes for working professionals who cannot attend regular classes. A training center could offer short courses to learners who want to acquire specific skills.

For institutes, a platform to start their own branded online academy in India can be the first step to this model. Depending on the features it offers it can be a tool for organizing courses, sharing learning materials, managing student access, and delivering educational content online.

Of course, starting an academy is more than just setting up the technology. Institutes need to think about what courses they will offer, who their clients will be and how they will offer a valuable learning experience. Course content should be clear, engaging and appropriate to the needs of the target audience.

It is equally important to consider what happens after a student enrolls. Easy access to lessons, clear instructions, timely communication and appropriate support for learners can make a real difference.

A successful online academy is a natural extension of what an institute already does well. Technology takes care of the teaching there, but the quality of the teaching is what gives the students a reason to stay.

Hybrid Learning: A Viable Option for Coaching Institutes?

Not every institute is prepared to go full online, and many students still enjoy the experience of learning in a classroom. Hybrid learning is the middle ground.

In simple terms, hybrid learning is a mixture of classroom teaching and online classes or digital learning resources. Students can attend regular classes at the institute and revise at home with recorded lectures or study materials online. Institutes may also offer selected sessions online to students who live farther away.

Take an example of a coaching center where some students come to class physically and others can’t travel regularly. By employing a hybrid model the institute can maintain service to both groups without completely reworking its existing instructional model.

Students enjoy greater flexibility, while teachers can tap into digital resources to enrich classroom instruction. Recorded lessons, when available, can be especially useful for students who need to review a difficult concept before an examination.

A platform for hybrid classes in coaching institutes can help to bring these learning activities in a more organized environment. Depending on the platform’s capabilities, institutes may be able to manage live sessions, recorded lessons, study materials and assessments along with their regular teaching activities.

The key is to ensure that online students receive the same amount of attention and clarity as those in the classroom. Timetables should be simple to follow, learning materials accessible and communication between teachers and students consistent.

Hybrid learning, if planned well, helps institutes retain the intimacy of classroom teaching, while giving students the convenience of digital education.

How an Institute Can Grow Through Digital Learning

Growth does not have to mean opening another branch or increasing the number of classroom batches. Sometimes it means finding new ways to transfer the knowledge and teaching expertise that an institute already has.

Digital learning can help make that happen.

An institute may introduce additional courses, provide recorded lessons for revision, or design specialized programs for students with different learning needs. It can also reach learners in locations where it would be difficult or expensive to establish a physical center.

There are practical benefits, as well. If learning materials and course activities are managed through a central platform, teachers can spend less time looking for resources or doing repetitive administrative work. Students find it easier to find the materials they need, and institutes can run their courses more consistently.

In addition, digital learning can assist institutions in adapting to changing student expectations. Some learners like live classes, others prefer recorded lessons and many are helped by a combination of both. An institute can reach a wider audience by providing a variety of learning formats.

But technology is only one part of the story. Institutes still need good teachers, relevant courses, good communication and a true understanding of their students. A digital platform can be of help, but it cannot substitute the trust that is built by quality teaching and a good learning experience.

ItLearn: Making Digital Learning Accessible

ItLearn by Imbibe Technologies is an education technology platform to assist institutions in delivering and managing digital learning. ItLearn offers an opportunity for coaching institutes who wants to go beyond traditional classroom teaching to explore a more structured approach to online education. It can support an institute’s efforts to make learning accessible to students beyond the physical institute and to adapt to evolving expectations around digital education. Institutions considering this transition can reach out to the ItLearn team to learn how the platform’s functionalities cater to their requirements for course delivery, learning materials, and student accessibility. No matter if the goal is to enhance an existing digital learning model or to discover new ways to deliver education, the first step is to determine what the institute needs and how ItLearn can fit into that plan.

Conclusion: Bring Your Institute Outside the Classroom

Going digital needn’t mean giving up on traditional teaching. That means more opportunities for your institution to reach students, to share knowledge, to fit learning more seamlessly into their lives.

Maybe you want to reach students in other cities. Or maybe you want to offer your own online courses. Or maybe you want to combine classroom and digital learning. Whatever your goal, the right platform will make the changeover easier to handle.

First look at the needs of your students and where your existing model of teaching could be more elastic. Then select a solution that supports those needs without unnecessarily complicating everyday teaching.

Perhaps your institute's growth will begin outside the classroom next. Get in touch with ItLearn from Imbibe Technologies today to talk about your digital learning needs and find out how you can take your coaching institute online, reach out to more students and create a future-ready learning experience.

Inside UniUI: the code and architecture behind a study platform for Nigerian universities

Fri, 09 Oct 2026 06:43:10 +0000

I'm a student at the Federal University of Technology, Owerri (FUTO). Over the last five months I've built UniUI, a platform where a Nigerian university student uploads their whole semester (courses, notes, past questions, deadlines, groups) and gets help running it. It's live at FUTO with 152 students and launches publicly in 50 days.

This post is the engineering tour: the data model, the two content pipelines, retrieval, the wallet, the agents that watch the platform, and the offline sync design, with the mistakes I made along the way.

A note on the code. Everything below is simplified and illustrative. I've kept real names where they're real (table names, tools, numbers) and removed secrets, edge cases and anything that would just add noise. Treat the snippets as sketches of the design, not copy-paste production code.

What's in it

2,000+ API routes and 237 database tables across 5 schemas
4,936 courses, 64,747 content chunks, 11,128 retrieval sections
11,071 structured study notes, 21,607 past questions, 3,708 formulas
A 10-stage upload pipeline and a 6-stage content pipeline
7 operations agents running as containers
20+ containers on one Oracle Cloud VPS with 7.8GB of RAM

Layer	What I use
Web	Next.js 15 (App Router), TypeScript, Tailwind, shadcn/ui, TanStack Query, Zustand, react-hook-form with zod
Mobile	React Native with Expo and Expo Router, SQLCipher for encrypted on-device storage
API	Python 3.12, FastAPI, Pydantic v2, asyncpg, SQLAlchemy, Alembic, Celery
Data	PostgreSQL 18 with pgvector, pg_cron, pg_trgm and pgcrypto; Supabase for identity; Redis; MinIO; MeiliSearch
AI	Voyage AI embeddings, OpenAI models in three tiers, Whisper, Tesseract, FFmpeg
Payments	A primary provider with Paystack and Monnify as fallbacks, NIBSS for bank transfers
Ops	Docker Compose, Caddy, Cloudflare, Sentry, Prometheus, Grafana, Loki, UptimeRobot

The shape of the system

Next.js web / React Native app
          |
        Caddy (TLS, routing)
          |
        FastAPI --- Pydantic validation on every request and response
          |
   +------+-----------------------------+
   |                                    |
 Postgres + pgvector              Celery workers
 Redis / MinIO / MeiliSearch      (uploads, notifications, batch jobs)
   |
 AI orchestrator -> Voyage embeddings + OpenAI tiers

Jarvis guardians -> core -> policy engine (OPA) -> audit ledger
(watching all of the above)

The rule I try to follow is that nothing crosses a boundary without a type. On the web that's TypeScript and zod. On the API it's Pydantic on every request, response and state object. That matters more than it sounds once agents are writing a lot of the code.

Data model first

Most of my early mistakes were data mistakes.

A course code doesn't identify a course. MTH 101 at one university isn't MTH 101 at another, so courses are unique per institution, not globally:

CREATE TABLE courses (
  id             bigserial PRIMARY KEY,
  institution_id bigint   NOT NULL,
  code           text     NOT NULL,   -- e.g. 'MTH201'
  title          text     NOT NULL,
  level          smallint,
  UNIQUE (institution_id, code)
);

Raw material becomes chunks. Chunks get grouped into retrieval_sections, and those are what the AI actually searches:

CREATE TABLE chunks (
  id          bigserial PRIMARY KEY,
  course_id   bigint REFERENCES courses(id),
  source_id   bigint NOT NULL,           -- the upload or document it came from
  ord         int    NOT NULL,
  body        text   NOT NULL,
  embedding   vector(1024),              -- null until embedded
  quarantined boolean NOT NULL DEFAULT false
);

CREATE TABLE retrieval_sections (
  id         bigserial PRIMARY KEY,
  course_id  bigint NOT NULL REFERENCES courses(id),
  topic      text,
  body       text   NOT NULL,
  source_ids bigint[] NOT NULL,          -- provenance: which sources fed this section
  embedding  vector(1024) NOT NULL
);

CREATE INDEX ON retrieval_sections USING hnsw (embedding vector_cosine_ops);
CREATE INDEX ON retrieval_sections (course_id);

Two details do real work here.

Provenance. Every chunk knows its course and source, and every section keeps the list of sources it was built from. The product promise is that an answer comes from your course. If I can't prove where a chunk came from, I can't keep that promise.

Only some things are embedded. Of the 64,747 raw chunks, about 15% are embedded. All 11,128 retrieval sections are. You embed what you search, and you search the curated layer.

The other data mistake was leaving structured LLM output in JSONB, which is quick to create and painful to query. I now normalise on write, and migrate old data like this:

CREATE TABLE section_topics (
  section_id bigint NOT NULL,
  topic      text   NOT NULL,
  difficulty smallint,
  PRIMARY KEY (section_id, topic)
);

INSERT INTO section_topics (section_id, topic, difficulty)
SELECT e.section_id, t->>'topic', (t->>'difficulty')::smallint
FROM section_enrichment e,
     jsonb_array_elements(e.payload->'topics') AS t
ON CONFLICT DO NOTHING;

JSONB is a fine place to land data and a bad place to leave it.

The upload pipeline (10 stages)

Students upload photos of notes, PDFs, and lecture recordings. Before any of it touches the corpus it goes through ten stages: client-side validation, chunked upload, virus scan, text extraction, classification, deduplication, quality scoring, course rep review, enrichment, and attribution.

Each stage is a Celery task, and the design rule that saved me the most pain is that every stage must be safe to run twice. Workers die, tasks retry, and an upload shouldn't be processed twice because a container restarted.

@shared_task(bind=True, acks_late=True,
             autoretry_for=(TransientError,), retry_backoff=True, max_retries=5)
def stage_extract(self, job_id: int) -> int:
    job = jobs.get(job_id)
    if job.stage_done('extract'):          # idempotent: a retry is a no-op
        return job_id
    text = EXTRACTORS[job.mime_type](job.object_key)   # pdf, image, audio, video
    jobs.save_stage(job_id, 'extract', output=text)
    return job_id

def start_pipeline(job_id: int) -> None:
    chain(stage_scan.s(job_id), stage_extract.s(), stage_classify.s(),
          stage_dedupe.s(), stage_score.s(), stage_route_review.s(),
          stage_enrich.s(), stage_attribute.s()).apply_async()

The upload_jobs table records every job, stage and error, so when a student says their upload vanished I can answer in one query. Extraction is mostly well-worn tools: pdftotext for PDFs, Tesseract for images, FFmpeg then Whisper for audio and video.

Deduplication

This matters in Nigeria in a way it might not elsewhere. The same past questions get forwarded across WhatsApp groups for years, re-photographed, re-typed, with a page missing. Exact hashes catch the lazy copies. MinHash catches the near-duplicates:

from datasketch import MinHash, MinHashLSH

def minhash(text: str, k: int = 5, num_perm: int = 128) -> MinHash:
    words = text.lower().split()
    m = MinHash(num_perm=num_perm)
    for i in range(len(words) - k + 1):
        m.update(' '.join(words[i:i + k]).encode())
    return m

lsh = MinHashLSH(threshold=0.85, num_perm=128)

def is_duplicate(job_id: str, text: str) -> bool:
    m = minhash(text)
    if lsh.query(m):          # something at least 85% similar already exists
        return True
    lsh.insert(job_id, m)
    return False

Quality scoring and human review

Every upload gets a 0 to 100 score on a six-factor rubric (legibility, completeness, course match, structure, originality, language). Routing is the generic part:

AUTO_APPROVE, REJECT_BELOW = 80, 40

def route(score: float) -> str:
    if score >= AUTO_APPROVE:
        return 'enrich'
    if score < REJECT_BELOW:
        return 'reject'
    return 'course_rep_review'     # a human decides

The middle band goes to a course rep, a student in the same course who approves or rejects. That human loop is deliberate: a rep knows when a past question paper is from the wrong year or the wrong lecturer, and a model doesn't.

The content pipeline (6 stages)

Stage	Input	Output
1. Chunking	Source material	64,747 chunks
2. Processing	Chunks	11,128 sections
3. Enrichment	Sections	11,098 enriched
4. Rich notes	Enriched sections	11,071 notes
5. Derived content	Rich notes	Practice questions, formulas, definitions
6. Indexing	Rich notes	Search index

The numbers shrink at each step, and that attrition is the quality gate doing its job. Another 1,951 chunks are quarantined entirely. I'd rather lose content than serve bad content.

To run stages over tens of thousands of rows I use the Postgres work-queue trick, FOR UPDATE SKIP LOCKED. Several workers claim batches without stepping on each other and without a separate queue:

WITH batch AS (
  SELECT id FROM chunks
  WHERE embedding IS NULL AND NOT quarantined
  ORDER BY id
  LIMIT 200
  FOR UPDATE SKIP LOCKED
)
SELECT c.id, c.body FROM chunks c JOIN batch USING (id);

Two things I'd flag for anyone doing this. Use input_type='document' when indexing and 'query' when searching, because the embedding model treats them differently. And do the whole claim-embed-write cycle in one transaction per batch, so a crashed worker releases its rows and the next one picks them up.

Retrieval and answers

The retrieval query is course-scoped first and ranked second:

-- pgvector 0.8+: keep scanning the HNSW index until the course filter is satisfied
SET LOCAL hnsw.iterative_scan = relaxed_order;

SELECT id, topic, body, source_ids,
       1 - (embedding <=> $1::vector) AS similarity
FROM retrieval_sections
WHERE course_id = $2
ORDER BY embedding <=> $1::vector
LIMIT 5;

That iterative_scan line is worth knowing. An HNSW index finds nearest neighbours across the whole table, then your WHERE course_id = ... filter throws most of them away, so you can get fewer than five rows back even though plenty exist. Iterative scans keep going until the filter is satisfied.

The orchestration is deliberately boring: embed the question, retrieve, refuse to guess if retrieval is weak, otherwise generate from the retrieved sections only.

class Answer(BaseModel):
    text: str
    sources: list[Source]
    confidence: float = Field(ge=0, le=1)
    grounded: bool          # False when nothing in the course matched well

async def ask(question, student_id, course_id, conn) -> Answer:
    qvec = vo.embed([question], model='voyage-3', input_type='query').embeddings[0]
    rows = await conn.fetch(RETRIEVAL_SQL, str(qvec), course_id)
    if not rows or rows[0]['similarity'] < MIN_SIMILARITY:
        return Answer(text="I couldn't find this in your course materials yet.",
                      sources=[], confidence=0.0, grounded=False)
    context = '\n\n'.join(f"[{r['id']}] {r['body']}" for r in rows)
    text = await llm_answer(model=MODEL_MINI, question=question, context=context)
    ...

The most important line is the early return. A study tool that confidently makes things up is worse than one that says it doesn't know, because a student will carry a wrong answer into an exam.

The Student Graph

Plain retrieval finds relevant paragraphs. What makes "study this, here's why" possible is knowing that this particular student keeps missing questions on one topic. I model that as a graph of 25 node types and 30 edge types (enrolled_in, mastered, struggles_with and so on), held in Postgres as a typed edge table:

CREATE TABLE graph_edges (
  src        bigint NOT NULL,
  dst        bigint NOT NULL,
  kind       text   NOT NULL,        -- 'struggles_with', 'mastered', 'enrolled_in', ...
  weight     real   NOT NULL DEFAULT 1,
  updated_at timestamptz NOT NULL DEFAULT now(),
  PRIMARY KEY (src, dst, kind)
);

Three model tiers

To control cost I route work through a nano model for classification and noise filtering, a mini model for daily work, and a larger model I only escalate to for critical incidents. Repeated queries go through a cache. The lesson is to test models on your own data. I ran the tiers against real student questions and real course material, and on my workload the cheapest model beat the expensive one. I'd never have guessed that from a leaderboard.

Plans and quotas

I sell AI usage by quota instead of by the month, which makes cost per student predictable:

Plan	Price	Allowance
Quick Ask	₦500	10 questions in 24 hours
Sprint	₦2,500	100 questions in 14 days
Scholar	₦2,500	120 questions in 30 days
Scholar Plus	₦3,500	250 questions in 30 days
Deep Study	₦10,000	Unlimited for 30 days, with offline access

Decrementing a quota must be atomic. Two simultaneous requests shouldn't both spend the last question, so let the database enforce it. No row back means no questions left:

UPDATE quotas
SET remaining = remaining - 1
WHERE student_id = $1 AND remaining > 0 AND expires_at > now()
RETURNING remaining;

I run the spend and the answer in one transaction, so a failed answer rolls the question back instead of charging a student for an error.

A ledger you can't quietly edit

The wallet is where bugs cost real money, so it's the part I was most careful with. Three rules:

Money is integer kobo. Never floats.
The ledger is append-only. Balances are computed, never stored and edited.
Every write carries an idempotency key, so a retried request can't pay twice.

CREATE TABLE wallet_ledger (
  id              bigserial PRIMARY KEY,
  wallet_id       bigint      NOT NULL,
  entry_type      text        NOT NULL,   -- one of 31 types, e.g. 'task_escrow_hold'
  amount_kobo     bigint      NOT NULL,   -- credits positive, debits negative
  idempotency_key text        NOT NULL UNIQUE,
  created_at      timestamptz NOT NULL DEFAULT now()
);

-- Make 'append-only' a property of the database, not a promise in the code.
CREATE FUNCTION forbid_ledger_changes() RETURNS trigger AS \(
BEGIN RAISE EXCEPTION 'wallet_ledger is append-only'; END;
\) LANGUAGE plpgsql;

CREATE TRIGGER ledger_no_update
  BEFORE UPDATE OR DELETE ON wallet_ledger
  FOR EACH ROW EXECUTE FUNCTION forbid_ledger_changes();

CREATE VIEW wallet_balances AS
SELECT wallet_id, SUM(amount_kobo) AS balance_kobo FROM wallet_ledger GROUP BY wallet_id;

If something is wrong, you don't edit a row. You append a reversal. That sounds annoying until the first time you have to explain exactly what happened to someone's ₦4,000.

A transfer writes both sides in one transaction and locks the source wallet first, so two simultaneous spends can't both pass the balance check:

async def transfer(conn, *, src, dst, amount_kobo, entry_type, key):
    async with conn.transaction():
        await conn.execute('SELECT 1 FROM wallets WHERE id = $1 FOR UPDATE', src)
        balance = await conn.fetchval(
            'SELECT COALESCE(SUM(amount_kobo), 0) FROM wallet_ledger WHERE wallet_id = $1', src)
        if balance < amount_kobo:
            raise InsufficientFunds()
        await conn.executemany(
            'INSERT INTO wallet_ledger (wallet_id, entry_type, amount_kobo, idempotency_key) '
            'VALUES ($1, $2, $3, $4) ON CONFLICT (idempotency_key) DO NOTHING',
            [(src, entry_type, -amount_kobo, key + ':debit'),
             (dst, entry_type,  amount_kobo, key + ':credit')])

Calling transfer twice with the same key is a no-op the second time. That removes a whole category of double-payment bugs caused by mobile networks that retry on timeouts.

Escrow as a state machine

UniUI has a marketplace, a task board, tutoring and a notes store, so money regularly sits between two students who've never met. I model every one as an explicit escrow state machine. If a transition isn't in the table, it can't happen:

ALLOWED = {
    S.HELD:      {S.COMPLETED, S.REFUNDED},
    S.COMPLETED: {S.RELEASED, S.DISPUTED},
    S.DISPUTED:  {S.RELEASED, S.REFUNDED},
    S.RELEASED:  set(),
    S.REFUNDED:  set(),
}

def move(escrow, to: EscrowState) -> None:
    if to not in ALLOWED[escrow.state]:
        raise IllegalTransition(f'{escrow.state} -> {to}')
    escrow.state = to

Task board fees, in integer kobo: the poster pays the reward plus 5%, the worker gets the reward minus 5%, and the platform keeps the 10% gap. For a ₦2,000 task, that's 210000 paid, 190000 received, 20000 kept.

Tutoring has a longer fuse. After a session, the tutor submits a note, both sides confirm attendance, and a 7-day dispute window opens before the money moves. A scheduled job releases whatever has aged out, again with SKIP LOCKED so workers can share the queue.

Payouts to bank accounts run in weekly batches. A wallet is only eligible with at least ₦10,000, a verified bank account, and an approved role (creator, ambassador, tutor and a few others), so the batch query encodes the rules directly.

Tokens have a price, and a cron job moves it

In-app rewards come as tokens with a naira value that floats between a floor of 40 kobo and a cap of ₦1.20. A pg_cron job reprices them every two hours, keeping the logic next to the data it reads. Prices are inserted, never updated, so there's a full history of what a token was worth at any moment.

Payment providers fail

The obvious fallback loop has a nasty trap. If provider A times out after actually charging the card, and you retry on provider B, you've charged the customer twice. So I only fall back on errors that prove nothing was charged, and I treat webhooks as the source of truth:

async def charge(providers, *, reference, amount_kobo, customer):
    last_error = None
    for provider in providers:
        try:
            return await provider.charge(reference=reference,
                                         amount_kobo=amount_kobo, customer=customer)
        except ProviderDown as err:      # raised only when we KNOW nothing was charged
            last_error = err
            continue
        # Any other exception (timeouts, ambiguous failures) is NOT retried here.
        # It goes to reconciliation, which asks the provider what really happened.
    raise PaymentUnavailable() from last_error

Webhook handlers must be idempotent too, because providers deliver the same event more than once:

CREATE TABLE webhook_events (
  provider    text NOT NULL,
  event_id    text NOT NULL,
  received_at timestamptz NOT NULL DEFAULT now(),
  PRIMARY KEY (provider, event_id)
);

-- If this returns no row, we've already processed the event.
INSERT INTO webhook_events (provider, event_id) VALUES ($1, $2)
ON CONFLICT DO NOTHING RETURNING event_id;

Referral attribution

Creators, ambassadors and student unions all earn from referrals, so they share one attribution system and one set of fraud rules. Attribution is last-click within 30 days:

SELECT c.referrer_id
FROM referral_clicks c
WHERE c.visitor_id = $1
  AND c.clicked_at <= $2
  AND c.clicked_at >= $2 - interval '30 days'
ORDER BY c.clicked_at DESC
LIMIT 1;

Commission is a percentage of the first payment plus a smaller percentage on recurring payments for six months, with every commission held for 7 days before it becomes payable and refunds reversing everything. The fraud checks are the unglamorous part: self-referral, duplicate signups from one device, click farming, and refund abuse. Even a crude version catches the cheap abuse.

Jarvis: agents with a policy gate

Separate from the agents that help me write code, I run a layer that watches the platform. I call it Jarvis. It has six guardian agents (infrastructure, security, users, growth, content, finance) and a persona called Treasure that's the only one allowed to talk to humans. Every agent is a container made of a YAML profile and a tools module.

What an agent produces is a typed signal, not free text:

class Signal(BaseModel):
    source: Literal['infrastructure', 'security', 'users',
                    'growth', 'content', 'finance']
    kind: str                                  # e.g. 'disk_pressure'
    severity: int = Field(ge=1, le=5)
    detail: dict
    proposed_action: str | None = None         # e.g. 'restart_container'
    target: str | None = None
    observed_at: datetime

The design decision I care about most is that the agent never decides whether it's allowed to act. The core orchestrator scores the signal, then asks Open Policy Agent, which lives in its own container outside any agent's reach:

package jarvis.actions

import rego.v1

default allow := false

protected := {"postgres", "jarvis-trust-ledger"}

# Low-risk restarts of ordinary containers run on their own.
allow if {
    input.action == "restart_container"
    not input.target in protected
    input.score >= 0.8
}

# Anything touching protected services asks a human first.
needs_approval if {
    input.action == "restart_container"
    input.target in protected
}

The orchestrator posts the action to OPA and gets back execute, ask_human (Treasure sends the approval request) or reject. Every decision, score and approval is written to an audit ledger in Postgres. Resolved incidents are stored in Qdrant with a signature, so the next similar incident can be matched against how the last one was fixed.

One distinction I try to keep honest: a cron job that restarts a container is automation. I only call something autonomy when the system decides what to do, does it, learns from the result and adjusts.

Device-first sync

Offline is the least finished part of the platform, so read this as design, not a victory lap. The app always writes to a local encrypted database first, and syncing is a background concern. Every local change also lands in an outbox:

-- On the device (SQLite, encrypted with SQLCipher)
CREATE TABLE outbox (
  seq          INTEGER PRIMARY KEY AUTOINCREMENT,
  entity       TEXT NOT NULL,
  entity_id    TEXT NOT NULL,
  op           TEXT NOT NULL,      -- 'upsert' | 'delete'
  payload      TEXT NOT NULL,      -- JSON
  base_version INTEGER NOT NULL,
  created_at   INTEGER NOT NULL
);

The sync loop pushes the outbox (deleting entries the server accepts, resolving conflicts on the rest), then pulls changes made elsewhere since its last cursor. Conflict strategy is per table: CRDT for notes, last-write-wins per field for goals, append-only for the journal.

For offline AI, the plan is quantised models on the device through llama.cpp, fetched from Hugging Face at download time so I don't store them on my own VPS, plus a limited WebLLM path in the browser.

Running 20+ containers on one 7.8GB VPS

Everything runs on a single Oracle Cloud VPS. It's cheap, which matters when you're a student, and it forces discipline about memory. The compose file is mostly memory limits, because without them one noisy container takes down the rest:

services:
  api:
    image: uniui-api-v2
    command: uvicorn app.main:app --host 0.0.0.0 --port 8000 --workers 4
    deploy: { resources: { limits: { memory: 1g } } }
    restart: unless-stopped
  worker:
    image: uniui-api-v2
    command: celery -A app.worker worker --concurrency 2
    deploy: { resources: { limits: { memory: 768m } } }
    restart: unless-stopped
  redis:
    image: redis:7
    command: redis-server --maxmemory 256mb --maxmemory-policy allkeys-lru
    restart: unless-stopped

Redis is split by database number so jobs can't trample each other's keys. Caddy terminates TLS in front, Cloudflare handles DNS and DDoS protection, Sentry catches errors, Prometheus and Grafana watch metrics, Loki collects logs, and UptimeRobot pings from outside.

Two infrastructure lessons. Uncapped Docker logs are a silent disk killer, fixed with max-size: 10m, max-file: 3. And MinIO's open-source project was archived and its images pulled from Docker Hub and Quay, so I had to move to the pgsty/minio fork.

Working with agents that write code

I'm one person, and agents wrote a large share of this code. That's the only reason it exists at this size, and it also created the problem I had to deal with. Agents produced 908 frontend pages, and I had to consolidate them to about 150 that matter for launch. Volume is cheap. Coherence is the part you still have to supply.

What I lean on:

A design contract. Every page uses components from one design reference file. If a pattern isn't there, I add it there first.
Types at every boundary. Strict TypeScript on the frontend, Pydantic on the backend.
Reversible migrations. Every schema change goes through Alembic, with a downgrade.
CI, which I'm adding. It's on my list and not done, and I'd rather say that plainly. The minimum gate is ruff, mypy --strict and pytest on the API, and tsc --noEmit plus lint on the web.

Here are the three incidents that shaped those rules.

Story 1: The agent that cried wolf 309 times

One agent ran a data-router audit. For every API route, it checked that the database column the route queries actually exists. It came back with 309 findings, all "broken column." I almost shipped the fixes.

Then I stopped. A working admin console doesn't have 309 broken columns. If it did, nothing would work. So I told the agent to verify its own work.

My code talks to two databases: Supabase for user data, and PostgreSQL for content. The agent had compared Supabase-backed routers against the content database's schema. profiles.email doesn't exist there, but it exists fine in Supabase. The audit was comparing apples to oranges and calling every difference a bug. After it checked against the right source of truth, the real count was 4.

What saved me was one line in the prompt: "If you're unsure whether to fix or delete, STOP and ask." An agent that stops to doubt itself is worth more than a confident one. Every agent now has a cry-wolf check: if a finding count is implausibly high, it must verify against the actual source of truth before reporting.

Story 2: The pages that lied

Early in the frontend build, an agent shipped a batch of pages. They compiled, the loading states worked, and the layouts matched the design system. They were also lying to students:

/likely-questions showed six exam questions with confidence scores and "appeared in 3/5 past years" labels. The agent had invented all of it. A student could have planned revision around questions no lecturer ever asked.
/departments listed departments with named HODs and student counts. None of it was real.
/verification shipped with status: 'Verified' as the default, so an unverified user saw a green tick.
/notes/create and /practice/flashcards/create discarded what the student typed, then redirected as if it had saved.

Fabricated data is the most dangerous kind of bug because it's invisible. There's no error and no crash, only a convincing lie. Three rules came out of it:

Never ship a page that lies. If the real action can't ship, remove the button. A missing button is honest.
No fabricated data, ever. Not for demos, not for placeholders.
Silent discard is worse than a crash. A crash tells the user something is wrong. A silent discard tells them their work is saved when it isn't.

Story 3: The four methods that weren't methods

A page showed a permanent loading spinner, with no error and no timeout. Four API client methods (getMyProgress, getPlatformBadges, followUser, redeemPromoCode) had been declared outside the ApiClient class instead of inside it:

// Inside the class: works
class ApiClient {
  async getMyProgress() { return this.request('/api/v1/my-progress'); }
}

// Declared outside the class: never becomes a method
async getMyProgress() { return this.request('/api/v1/my-progress'); }

TypeScript accepted it because the class had a [key: string]: any index signature. The callers used optional chaining:

apiClient.getMyProgress?.().then(...)

The method was undefined, so ?. short-circuited, .then() never fired, and the spinner never cleared. Optional chaining hid a real failure. The tests passed because none covered those pages, and lint had nothing to flag. Agents now scan for methods declared outside their class and for ?.() calls to methods that don't exist on the actual object.

How I review agent output

Every change goes through four gates:

Does it compile? Types, lint and tests. Table stakes, and it catches almost nothing important.
Does it actually work? I run it in the real environment: open the page, click the button, check the network tab and console. Most bugs live here.
Does it lie? Is every number real? Does every button do what its label says? Does every "saved" mean saved?
Does it break something else? What other pages call this code, what other routes depend on this schema, and what happens if it runs twice?

What I won't let an agent touch

Code. Not one line. Jarvis's guardians can watch containers, correlate logs, ban IPs, score content, and flag wallet anomalies. They cannot edit code, change a config file, modify the frontend, or deploy anything.

A bad code change has an unbounded blast radius. A subtle bug in the payment path costs real money, and a frontend regression hits every student at once. Recovery is slow and the failure modes are catastrophic, so even approval gates at every step don't make the risk-reward work.

Instead, Jarvis detects the bug, finds the root cause, drafts the fix, writes a test and opens a PR. Then I review it, approve it and deploy it. The agent does the analysis and I do the commitment. The operations layer earns trust through demonstrated performance. The code layer doesn't get to earn it yet, because I'd be trusting an agent to act correctly in situations I haven't imagined, at 3 AM, while I'm asleep.

What I'd tell someone starting today

Attach provenance to everything on day one. Course, level and source on every chunk. It makes every later promise possible.
Normalise on write, not on read. JSONB is a fine landing zone and a bad home.
Build the boring parts first. Auth, billing, notifications and sync took about 60% of my engineering time.
Test models against your data, not benchmarks. The cheapest one won on my workload.
Make money paths boring and strict. Integer kobo, append-only ledger, idempotency keys, explicit state machines.
Volume is cheap, coherence is not. Agents produce things. Keeping them consistent is your job.
Some blockers aren't code. My mobile app, with 152 screens built, sat blocked for weeks on an Apple Developer account.

What's next

Public launch is in 50 days, with a planned expansion to six more Nigerian universities in November 2026. I'm recruiting 500 student creators to help distribute it, and the unglamorous legal work (CAC registration, NDPA compliance, a proper privacy policy) has to land first.

If you've built retrieval over messy educational content, wallets or escrow in Nigeria, or a product with a lot of agent-written code, I'd like to hear what you'd do differently. Tell me in the comments.

App: app.uniui.com.ng
Waitlist and creator program: waitlist.uniui.com.ng
Main site: uniui.com.ng

Why Machine Learning Is Hard to Learn — And How Interactive Learning Can Help

Mon, 05 Oct 2026 11:34:36 +0000

Machine learning is one of those subjects that looks relatively simple from a distance.

You train a model, give it some data, and it makes predictions.

Then you start learning how it actually works.

Suddenly you're dealing with gradients, loss functions, optimization, vectors, embeddings, tokenization, attention, and a long list of other concepts.

The difficult part isn't always the mathematics.

A big part of the problem is that many machine learning concepts are invisible.

You can't see a gradient moving through a neural network. You can't directly see an embedding space. You can't watch attention weights changing inside a transformer just by reading a definition.

This is where interactive learning can make a difference.

Reading about something isn't the same as understanding it

Consider gradient descent.

A typical explanation might say:

Gradient descent is an optimization algorithm that iteratively adjusts parameters in the direction that minimizes a loss function.

That's technically correct.

But for someone learning machine learning, it can still be difficult to build an intuition for what is actually happening.

You have a function.

You have parameters.

You have a loss.

And somehow the parameters are updated repeatedly until the model reaches a better solution.

An interactive visualization can make this much easier to understand.

Instead of only reading the explanation, you can change a parameter and observe what happens to the loss.

The concept becomes something you can experiment with.

Visualization creates a mental model

Good visualizations aren't just decoration.

They can help create a mental model.

For example, imagine learning a simple regression model.

You could see:

The training data
The current prediction line
The error between predictions and observations
The loss changing over time
The parameters being updated

Now the optimization process becomes a sequence of observable events rather than a collection of formulas.

This doesn't eliminate the mathematics.

It gives the mathematics something concrete to attach itself to.

The same problem exists with neural networks

Neural networks introduce another layer of abstraction.

You have inputs, weights, biases, activations, layers, and eventually an output.

When you're learning the subject for the first time, it's easy to memorize the terminology without really understanding how the pieces interact.

An interactive neuron can help.

Change an input.

Change a weight.

Observe the output.

Then add another neuron.

Then another layer.

The architecture starts becoming less mysterious.

LLMs make this even harder

Modern language models introduce another collection of concepts.

Tokenization is one example.

A sentence isn't simply passed to a language model as plain text.

Text is converted into tokens, which are represented numerically and eventually processed as vectors.

Then you encounter embeddings, attention, positional information, training objectives, instruction tuning, and inference.

Each individual concept is manageable.

The challenge is understanding how they connect.

That's why I think learning the pieces in the right order is important.

Start with the fundamentals

It's tempting to jump directly into transformers because they're everywhere.

But understanding some of the concepts underneath them can make the architecture much easier to follow.

A reasonable progression might look something like:

Linear models → neurons → backpropagation → vectors → tokenization → attention → training → fine-tuning → inference

You don't necessarily need to become an expert in every stage before moving forward.

But having a basic mental model of each stage makes the next one easier.

Interactive learning doesn't replace traditional learning

This doesn't mean textbooks, courses, documentation, or research papers are no longer useful.

They're extremely useful.

Interactive learning works best as another layer.

A useful learning process might look like:

Explanation → visualization → experimentation → deeper reading

First understand the basic idea.

Then interact with it.

Then experiment.

Once the concept makes sense, the more mathematical explanation becomes easier to digest.

I wanted to experiment with this approach

This was one of the reasons I started building MasterAI Now, an interactive learning platform for machine learning and LLM concepts.

The idea is to combine short explanations with interactive playgrounds and knowledge checks.

For example, instead of explaining a concept entirely through text, a lesson can introduce the idea and then let the learner experiment with a simplified version of it.

You can explore the approach at MasterAI Now.

The project currently follows a structured path from fundamental machine learning concepts toward modern LLM topics such as self-attention, LoRA, RLHF/DPO, and inference.

The important part is curiosity

One thing I've noticed while working on educational material is that experimentation changes the learning process.

When you can change something yourself, you naturally start asking questions.

"What happens if I increase this?"

"Why did the loss increase?"

"What happens if I remove this?"

"Why does the model behave differently now?"

Those questions are valuable.

They turn learning from passive consumption into exploration.

Final thoughts

Machine learning is difficult partly because so much of it happens at an abstract level.

You read about a process that you can't directly observe, memorize some terminology, and then try to connect everything together.

Interactive learning can reduce that gap.

It doesn't make machine learning magically easy.

It simply gives you another way to approach the same concepts.

Read the explanation.

See the process.

Change something.

Observe the result.

Then go deeper.

For me, that's a much more interesting way to learn how AI actually works.

MasterAI Now

Mon, 05 Oct 2026 11:31:47 +0000

MasterAI Now — Learning AI by Interacting With It

Learning how modern AI works can be difficult.

There are countless tutorials, courses, blog posts, and research papers available, but I kept running into the same problem: reading about a concept doesn't always make it intuitive.

You can read about backpropagation for an hour and still have trouble visualizing what is actually happening.

I wanted to build something that approached the problem differently.

That's how MasterAI Now started.

The idea

The goal is to make machine learning and LLM concepts easier to understand through a combination of short explanations, interactive simulations, and knowledge checks.

Instead of treating AI education as a collection of long articles, the learning experience is structured as a progression.

You learn a concept, interact with it, and then check whether you understood it.

The idea is simple:

Learn → interact → test → continue.

What can you learn?

The current learning path follows important concepts across the machine learning and modern LLM lifecycle.

It currently covers topics including:

Linear Regression
Artificial Neurons
Backpropagation
Data Deduplication with LSH
BPE Tokenization
Vector Spaces
Self-Attention
Pre-training and Loss
Instruction Tuning with LoRA
Alignment with RLHF and DPO
Inference and KV-Cache

The progression is intentional.

Rather than starting with transformers and assuming the reader already understands everything underneath them, the material starts with simpler concepts and gradually builds toward modern LLM architecture and inference.

Interactive playgrounds

The part I'm most interested in is the interactive side.

Some AI concepts are difficult to understand from text alone.

For example, you can read an explanation of gradient descent, but being able to change parameters and see how the model moves toward a better solution creates a different learning experience.

MasterAI Now includes interactive playgrounds designed around the concepts being taught.

They aren't intended to replace production machine learning frameworks.

They're educational simulations designed to make the underlying mechanics easier to see.

Short learning cards

The lessons are broken down into smaller sections instead of presenting everything as one long article.

Each section focuses on a specific idea and tries to explain it without unnecessary jargon.

The goal is to make the material approachable for someone who understands basic programming but doesn't necessarily have a machine learning background.

After the explanations, knowledge checks help reinforce the concepts before moving to the next stage.

Who is it for?

MasterAI Now is mainly aimed at developers, students, and curious builders who want to understand what's happening underneath modern AI systems.

You don't need to be an ML researcher to start.

You also don't need to immediately dive into complex research papers.

The goal is to provide a practical learning path between knowing how to use AI products and understanding the concepts that make them work.

Why I built it

There are already excellent resources for learning machine learning.

This project isn't trying to replace them.

I wanted to experiment with a different learning format where concepts are introduced progressively and supported by interactive visualizations.

Building the platform has also taught me something interesting:

Educational software isn't only about presenting correct information.

The order in which information is introduced matters just as much.

If a concept is introduced too early, it can become confusing. If it's introduced too late, previous concepts may not make sense.

Finding that balance has been one of the most interesting parts of building the project.

What's next?

The core learning path is now available, but there is still plenty to improve.

Some areas I'm working on include:

More interactive simulations
Better visual explanations
Additional AI and ML topics
Improvements based on learner feedback
More practical examples
A clearer progression between advanced concepts

The project is currently free to use.

You can explore it here:

https://masterainow.com

If you're interested in AI, machine learning, or understanding how LLMs work under the hood, I'd be interested to hear what concepts you found hardest to understand when you first started learning them.

Too Much Manual Work After Online Training? Automate Testing and Certification

Thu, 01 Oct 2026 09:56:30 +0000

You finish an online training. Everyone attended, the session went fine, and you feel good about it then the real work starts. Someone has to make the test papers, someone has to check the answers, someone has to make the certificates one by one and send them by mail.

If you run training for your company, school, or institute, you know this pain very well. The training takes two hours, but the follow-up work takes two days and most of the time, it is the same work again & again.

There is an easier way, you can let a system do the testing and certification for you. This article will tell you how, and what to look for. If you are searching for the best platform for online courses with certification and testing, keep reading, because the choice you make here saves you a lot of time.

Why Manual Work Becomes a Big Problem

At the start, manual work feels okay. You have 20 learners, so you make a Google Form, check the marks by hand, and type names on certificates. It takes some time, but you manage.

Then the number grows, now there are 100 learners, then 500. The same method does not work anymore- checking papers takes days, mistakes start coming in, a name is spelled wrong on a certificate, somebody's marks are missed and someone gets a certificate who did not even pass.

And the learners are waiting. They finished the course last week and still have no result, they keep asking, "When will I get my certificate?" You feel bad, but you are also busy with a hundred other things.

This is the main problem with manual work, it does not grow with you. The more you grow, the more tired your team gets.

What Can Be Automated

Most people think only the exam checking can be automated but actually many things can be done by the system on its own.

First is the test itself. You make the questions once and the system gives them to every learner, it can mix the questions, set a time limit, and check the answers as soon as the learner submits. There is no need to sit with a pen and check papers.

Second is the result, the learner sees the marks right away. If they pass, good. If not, they can try again, if you allow it.

Third is the certificate. When a learner passes, the system makes the certificate with their name and course details and sends it to them- no typing, no copy paste, no waiting.

Fourth is record keeping- who joined, who finished, who passed, who failed, everything is saved in one place. When your boss or a client asks for a report, you just open it and show.

Saving Time and Reducing Mistakes

Think about how much time your team spends on this work each month. Even if it is just ten hours, that is a lot of time that could go into better things, like making better courses or helping learners who are stuck.

Automation also cuts down mistakes. Humans get tired, when you check 200 papers, by the 150th one your eyes are already slow. A system does not get tired, it checks the 200th paper like the first one.

Certificates are another thing. When you make them manually, small errors happen like wrong name, wrong date, wrong course title. In a system, the details come directly from the learner's profile, so these mistakes almost go away.

There is one more benefit, learners trust you more. When they get the result and certificate the same day, they feel the training was serious and well organised. That matters, especially if the certificate is useful for their job.

What to Look For in a Platform

Not every tool is good. Some only host videos, some only do tests or some make certificates but with poor design, so before you pick one, check a few simple things.

The platform should let you run courses, tests, and certificates all in one place. If you use three different tools, you are still doing manual work to connect them.

It should be easy to use. Your team should not need a long training to learn the tool, if it takes weeks to understand, it defeats the purpose.

Test options matter too- look for different question types, timers, auto checking, and the choice to allow retakes. Also see if the platform can stop cheating in some way, like showing questions in random order.

Certificates should look nice, carry your own name or logo, and go out automatically. Also, reports should be clear, so you can see who is doing well and who is not.

Last, check the support. When something goes wrong at night before a big exam, you want a team that replies quickly.

Who Needs This Platform Most

Many kinds of people face this problem. A company that trains new employees every month needs it, a coaching institute that runs batches all year needs it, a trainer who sells online courses needs it, even schools and colleges that take short skill courses can use it.

If you are doing the same training again and again, and you hand-check every test, then you are the right person for automation. Even if your numbers are small today, it is better to set it up early. Changing later, when you have thousands of learners, is much harder.

How to Start Without Stress

Many people delay because they think the change will be heavy. It is not, you can start small.

Pick one course. Put your content on the platform. Make a short test with ten or fifteen questions. Set the passing marks. Add your certificate design then invite a small group of learners and see how it goes.

When it works well, move the rest of your courses one by one. Don't try to shift everything in one day, slow and steady is fine.

Also take feedback from your learners. Ask them if the test was easy to open, if the certificate came on time and if anything was confusing, their answers will help you fix small things early.

Final Words: Stop the Manual Work Today

Training is meant to help people learn, not to keep you buried in checking papers and typing names. Your time is better used on teaching and growing your business.

The right system does the boring work for you. Tests get checked on their own, certificates go out on their own, you get clean reports without any extra effort. Your learners are happy, and your team can breathe.

If you want the best platform for online courses with certification and testing then ItLearn by Imbibe Tech is a good place to start, it is built so that you can run your course, take the test, and give the certificate from one single place. No more switching between tools, no more late nights checking papers.

Choose ItLearn by Imbibe Tech and see how much easier your training work becomes. It is the best platform for online courses with certification and testing for trainers, companies, and institutes who are tired of manual work.

Don't wait for the next batch to pile up. Contact today, set up your first course, and let the platform for online courses take care of the rest.

Where Habit Meets Understanding

Mon, 28 Sep 2026 04:36:15 +0000

Last week I told you about Chapters 1-3. This week I moved on to Chapters 4-6.

The terminal is not new to me, but going back to the fundamentals on purpose is teaching me things I'd been doing on autopilot 😂

Let me take you through how the week went.

It Started With Open Source

I've heard "open source" a million times and nodded like I fully understood. This week I finally got into what it really means: the code is out there for anyone to read, use, change and share.

Then the acronyms arrived 😭 GPL, copyleft, FOSS, FSF, OSI, Creative Commons. My brain was like, "Sis, all of this in one chapter?"

What helped was slowing down and asking what each one is for:

FOSS is free and open source software
GPL is a license that keeps software free, even after people modify it
Copyleft is the idea behind that: if you build on free software, you keep it free
FSF and OSI are the organisations that push and define all of this
Creative Commons does a similar job for creative work like writing, images and music

Tip: Don't memorise acronyms. Understand the purpose and the names will stick on their own.

Then Came the Command Line

I did the labs and typed everything out myself instead of just reading, and chale, going back to basics on purpose hits different. Some of these commands I use often, but this time I actually looked at what each one does:

echo      # displays text or values
history   # shows commands you've used before
type      # tells you what kind of command something is
which     # shows where a command lives
man       # opens the manual for a command

Tip: history is underrated. It's a record of everything you've tried, mistakes included 😂

Chapter 6 Taught Me What Matters Most

This chapter reminded me of something every good hacker knows: you don't have to know everything, you just need to know how to find help.

Here's the toolkit:

man ls      # read the manual for a command
whatis      # quick one-line description
man -k      # search manuals by keyword
apropos     # same idea as man -k
whereis     # find a command and its documentation
locate      # find files and directories
info        # more detailed documentation
--help      # quick help straight from the terminal

Tip: Before you open Google, try --help or man first. It trains you to find answers on your own, and that's a real skill.

The Part That's Still Hard

The vocabulary. Sometimes I'm reading and I stop like, "Wait, what does this one mean again?" 😭

But I'm learning not to rush. I read, practice, get confused, check the documentation, try again, and somewhere in that loop it starts to click.

What I'm Taking Away

Linux isn't about memorizing every command. It's about understanding what you're doing and knowing how to find your way when you don't know something.

That's a lesson that goes beyond Linux, honestly.

The Duchess of Hackers
Full-Stack Developer | Digital Marketer | Cybersecurity Enthusiast

How to Identify Learning Gaps with Online Education Platforms to Track Performance

Fri, 25 Sep 2026 06:13:30 +0000

Every teacher has had this moment. You finish a chapter, class seems fine, nodding along, no questions, then the test comes back and half the class messes up the same three questions. What happened there? Where did it go wrong exactly?

That's a learning gap. Basically the space between what a student should know and what they actually know, spotting this early is genuinely one of the hardest parts of teaching. You've got 40 kids in a room, you can't sit with each one and check line by line what they've actually understood.

This is where tech steps in, more schools now are figuring out how teachers can use online education platforms to track performance properly, not just for marks on a report card. But to actually find where a student is stuck, and fix it before it snowballs into something bigger.

What Is a Learning Gap, Really?

It's not just a low score, it's more specific. A missing piece somewhere in the understanding, even if the student moved on to the next topic like everything was fine.

Say a kid didn't fully get fractions in class 5. Fast forward to class 7, ratios become a struggle. By class 9, algebra feels impossible, the root of it all goes way back, nobody caught it at the time though.

Happens all the time in regular classrooms, honestly. The syllabus is long, teachers move fast, there's no other option really, students who didn't get it stay quiet. Nobody wants to look slow in front of the whole class, so the gap just sits there- quietly growing, year after year.

Regular tests catch this too late usually, by the time a big exam shows the problem, months are already gone. This is exactly why schools are looking into how teachers can use online education platforms to track performance — something closer, more regular, not just twice a year.

Why Normal Classroom Methods Miss These Gaps

Let's be real. One teacher, forty students, forty minutes on the clock, there's no way to check everyone properly at that time. The teacher asks "any doubts?" Silence, doesn't mean there are no doubts, just means nobody's raising a hand.

Exams happen once every few months. By then a small gap from three months ago has already turned into three or four connected gaps. Student's now behind in multiple places, not just the one spot where it started.

Homework checking isn't perfect either. A teacher sees right or wrong answers but not always why. Silly mistake? Or an actual gap in understanding? Hard to tell from a notebook alone, most of the time.

This is the real issue with old-school teaching, not the teaching part — that's fine. It's the tracking that falls short and that's exactly the space online platforms are built to fill.

How Online Platforms Actually Catch These Gaps

Online platforms work differently. Every quiz, every assignment, every little practice question — all of it gets recorded, not just the final score. The specific question, the specific concept the student got wrong.

Say ten students in a class mess up a question but not always the same one. A good platform groups these errors by topic instead, so if seven out of forty keep messing up questions on "photosynthesis reactions" specifically, that's not random anymore. That's a real, visible gap sitting right there.

This is basically the answer to how teachers can use online education platforms to track performance in real life, not replacing the teacher — just giving actual data instead of guesswork. Instead of assuming the class got it, teachers can see exactly where things fell apart.

Some platforms even track time spent per question. A student who rushes through in ten seconds tells a very different story than one who spent five minutes and still got it wrong- both matter and both point to different problems entirely.

Tracking Each Student Not Just the Whole Class

Class-level data is useful, sure but individual tracking is where things actually get valuable. Every kid learns differently — some slow and steady, some fast but careless, a platform shows these patterns clearly, over weeks, not just from one test.

The teacher opens a student's profile, sees the whole journey laid out- strong topics, weak topics, the ones they keep failing at again and again. This kind of pattern is nearly impossible to remember on your own, especially juggling a full class of thirty-plus kids.

This individual view is a big chunk of how teachers can use online education platforms to track performance for each child separately, not just as one big average number. It's a completely different way of seeing the classroom.

Parents benefit too, actually. Instead of a vague "your child needs to improve," teachers can point to the exact topic. That kind of clarity builds trust and makes those awkward parent-teacher meetings a lot less awkward.

Turning the Data Into Actual Fixes

Finding the gap is only half the job, though. The real value is what happens after, good platforms don't just show the problem — they suggest what to do next, like extra practice on weak topics, revision material sent out automatically, alerts when a student keeps falling behind.

This turns tracking into something you can actually act on. A teacher doesn't sit there manually making twenty different worksheets for twenty different kids. The system already knows who needs what, based on real numbers collected over time, not a hunch.

This is really the whole point behind how teachers can use online education platforms to track performance in a way that's actually useful, not just numbers sitting on a dashboard nobody opens. A real step that helps a kid catch up before the gap turns into a wall.

Teachers can also use this to plan the next class better. Most of the class is weak in one area? That becomes tomorrow's focus- no guessing involved, no wasted time re-teaching stuff they already know.

Why This Matters Even More Today

Classrooms are more mixed than they used to be, some kids grasp things fast, others need more time and more repetition. One-size-fits-all teaching just doesn't hold up anymore, not with this much range in learning speed sitting in one room.

Online tracking tools aren't replacing teachers, to be clear, they just give teachers better eyes. Instead of guessing who's struggling, they know. Instead of waiting for the term exam to find out, they see it play out week by week, in real time.

This is exactly why more schools are getting serious about how teachers can use online education platforms to track performance — not as some extra add-on, but as part of how teaching actually works now, every day.

Final Thoughts

Learning gaps are normal, every student has one somewhere, in some subject. The real problem was never that gaps exist, it's something that nobody notices until it's too late to fix easily.

Understanding how teachers can use online education platforms to track performance changes this whole picture, moving teaching from guesswork to actual insight. From reacting late to catching things early- that difference matters a lot, especially for a student's confidence over time.

Try ItLearn by Imbibe Tech

If your school actually wants to understand online education platforms to track performance, ItLearn by Imbibe Tech is built exactly for this, helps teachers catch learning gaps early, track each student's progress, and act on real data instead of gut feeling.

This makes it easy for schools to finally see how teachers can use online education platforms to track performance in a way that actually helps kids, not just generates reports nobody bothers reading.

Don't let learning gaps quietly pile up in your classroom. Try today and see online education platforms to track performance the right way, before the small gaps turn into big ones.

It's Been a While - So Let's Talk About Going Deeper Into Linux

Thu, 24 Sep 2026 21:11:55 +0000

It's been a while since I last posted here life, work, and about a dozen half-finished drafts got in the way. But I'm back, and this time I have something worth sharing.

If you know me, you know I'm a cybersecurity person at heart, so Linux was never really a stranger to me. I've touched it, used it, worked around it enough to get by. But "getting by" and actually understanding it are two very different things and last week, I decided to close that gap properly.

So with my few knowledge in Linux going in, I went a bit deeper Linux Essentials, chapters 1 to 3. Not just using it anymore, but actually understanding what's happening underneath. Here's what that week taught me 👇

1. Linux Is a Kernel, Not an OS

This is the one that reframed everything, even with prior exposure.

Linux itself is just the kernel — it talks to hardware, manages memory and processes
Ubuntu, Debian, Fedora — those are distributions: the kernel + tools + a package manager, bundled for humans
Different distro, same kernel underneath

Tip: Once this clicks, a lot of the "why are there so many versions of Linux" confusion disappears.

2. Revisiting the Terminal, Properly This Time

I'd used a terminal before, but mostly copy-pasting commands without really knowing what they did. This time I slowed down:

pwd     # where am I
ls      # what's here
cd      # move around
mkdir   # make a folder
touch   # make a file
cat     # read a file

Tip: There's a difference between running a command because a tutorial told you to, and running it because you know exactly what it'll do. That's the gap I was closing.

3. Package Managers Aren't All the Same

Debian-based (Ubuntu, Mint) → .deb packages via apt / dpkg
RPM-based (Fedora, RHEL) → .rpm packages via yum / rpm

Tip: A package manager's real job isn't "install this" — it's quietly handling dependencies so what you install actually runs.

4. Security Starts Before You Open a Security Tool

Coming from cybersecurity, this chapter felt like home — but it still reframed a few things:

Root privileges — know who has them and why
Strong passwords + 2FA
Permissions done right
Ubuntu's UFW for firewalls (Gufw if you want it with a GUI)

Tip: Good security is a habit you practice, not a tool you install once.

5. Virtualization — The Concept Behind Half of Modern Tech

One physical machine (host) runs several virtual machines (guests)
A hypervisor manages them
Same underlying instinct behind Docker and Kubernetes — don't waste hardware on one job

Tip: This is exactly why you run tools in a VM instead of your main OS — a lesson every cybersecurity person learns the hard way eventually.

The Real Takeaway

The concepts weren't the hard part for me — the vocabulary review was. Kernel, shell, hypervisor, dependency, distro. Going back to the fundamentals as someone who already "knew" Linux was humbling in a good way. It's easy to skip the basics when you can already get things done. Going back and actually understanding why things work made all the difference.

18 chapters and a bunch of labs still ahead. I'll be documenting this properly from here — so consider this Part 1.

Stay curious, stay consistent

The Duchess of Hackers
Full-Stack Developer | Digital Marketer | Cybersecurity Enthusiast

The AI Era: Democratizing the Ivory Tower and Reclaiming the Latin American Mind

Wed, 09 Sep 2026 13:24:55 +0000

When you view the academic, professional, and technological worlds through the intersectional lens of a woman born and raised in the Global South, you quickly learn to look at every first-world narrative with absolute skepticism. For decades, the intellectual establishment of the global North has built massive, impenetrable walls around knowledge, dictating who gets to speak, who gets to write, and who counts as an authority. They have constructed an intricate system of gatekeeping designed to protect their prestige, maximize their fame, and extract the raw intellectual and natural resources of the rest of the world while leaving the creators in the footnotes.

For a long time, I lived with an intense, suffocating sense of self-doubt. I questioned whether my perspective was relevant, whether my observations had value, and even whether I had the right to claim the labels that describe my own existence. It is a common crisis of identity for those of us from Brazil. We trap ourselves in a bizarre linguistic limbo, wondering: Can I truly call myself Latina? Because first-world media and American-centric bureaucratic boxes heavily dominate global culture, the world has been conditioned to treat "Latino" as a concept synonymous only with Spanish speakers.

But this is a superficial, historically illiterate flattening of our reality. The very word "Latino" traces its lineage directly back to the Latin language of the Roman Empire, the root from which all Romance languages evolved. Spanish is a Latin-derived tongue, but so is Portuguese. To deny a Brazilian woman her identity as a Latina is to succumb to a geographic and administrative distortion invented by first-world institutions to simplify their demographics. Brazil is the largest, most populous, and geographically dominant nation in Latin America. When we claim the title of Latina, we are not borrowing a Spanish word; we are reclaiming our geography, our history, our language’s Roman lineage, and our collective place in the historic struggle against institutional exclusion.

The fact that we feel we must ask permission to use a word that defines our own home continent is the ultimate proof of how deeply the gatekeepers have succeeded. They have not only monopolized international publishing, funding loops, and prestigious prizes; they have colonized our internal sense of worth. When you look at the history of science and medicine, you realize this pattern of systemic extraction and erasure is centuries old.

The Footnotes of Empire: Ferreira, Chagas, and Finlay

The history of first-world academic prestige is a history of extraction. They take our raw data, our clinical observations, and our natural resources, process them through the rigid machinery of Western institutions, and brand the final product with a first-world trademark.

Consider the tragic case of Dr. Sérgio Henrique Ferreira. In the 1960s, working out of the University of São Paulo in Ribeirão Preto, Dr. Ferreira conducted brilliant, foundational research on the venom of the Bothrops jararaca pit viper. He succeeded in isolating the Bradykinin-Potentiating Factor (BPF), a discovery that unlocked the molecular mechanics of blood pressure regulation. This exact molecule became the indispensable biochemical blueprint for Captopril, the world’s very first ACE inhibitor, which went on to revolutionize cardiology and save millions of lives across the globe.

Yet, when Ferreira traveled to London to collaborate with his colleague Sir John Vane, co-authoring pioneering papers that mapped how aspirin inhibits prostaglandins to treat pain, the historical script played out precisely as written. In 1982, John Vane was awarded the Nobel Prize in Physiology or Medicine for those exact discoveries. The British supervisor received immortal glory, historical statues, and global fame, while the brilliant Brazilian mind whose hands-on expertise and venom-derived mechanics fueled the breakthrough was left out of the spotlight, transformed into an invisible column in an institutional ledger.

This pattern repeats across what the West patronizingly labeled "tropical medicine"—a term historically used to ghettoize the medical brilliance of the Global South as exotic and localized, rather than universal. In 1909, working inside a modified railway car in the interior of Minas Gerais, Carlos Chagas achieved a medical triumph that no European or American scientist had ever accomplished before, and none has duplicated since. He single-handedly discovered, isolated, and fully described a completely new infectious disease. In a single, historic sweep of independent execution, Chagas identified the parasite (Trypanosoma cruzi), the insect vector (the barbeiro), the domestic animal hosts, and the entire clinical manifestation of the disease that now bears his name. He was nominated for the Nobel Prize in 1913 and 1921. He never won. The Eurocentric committees simply could not stomach awarding the highest badge of scientific immortality to a Brazilian doctor working with a shoestring budget out of a train car.

We see the exact same corporate and academic gaslighting in the story of Cuban physician Carlos Juan Finlay. In 1881, Finlay discovered that yellow fever was not an airborne miasma, but a vector-borne disease transmitted specifically by the Aedes aegypti mosquito. For two decades, the elite medical establishments of Europe and the United States openly mocked him, dismissing his vector theory as an unscientific, backward colonial delusion. Decades later, when the U.S. Army Yellow Fever Commission led by Walter Reed arrived in Havana, they used Finlay's exact data, adopted his exact mosquito-breeding methods, confirmed his thesis, and claimed the absolute credit. Today, first-world textbooks worship Walter Reed as the heroic conqueror of yellow fever, while Finlay is pushed into the shadows of the historical narrative.

Worshiping the Gods of Condescension: Darwin and Einstein

We are trained to revere these institutions and their foundational giants, yet if you open the private notes of the men whose statues decorate first-world universities, their unfiltered colonial disdain for our people becomes glaringly obvious.

Take Charles Darwin, who is canonized in first-world textbooks as a saint of progressive thought and a fierce abolitionist who detested slavery. It is true that he wrote with absolute horror about the human atrocities he witnessed enslaved people suffering in Rio de Janeiro and Pernambuco. But academia completely hides his secondary worldview. When you open his private diaries from his time in Rio, specifically his entry on July 3, 1832, his Eurocentric, aristocratic bias is completely unfiltered. He wrote that the local Brazilians were:

“Ignorant, cowardly, & indolent in the extreme... contented with themselves & their customs, they answer all remarks by asking ‘why cannot we do as our grandfathers before us did’.”

He dismissed an entire population as naturally lazy and backward simply because they did not live, think, or perform according to 19th-century British standards.
Fast forward nearly a century later to May 1925, and you see the exact same toxic elitism from Albert Einstein during his week-long visit to Rio de Janeiro. To the Brazilian public, his visit was a monumental event; local scientists, doctors, and politicians treated him like a visiting deity, packing lecture halls and hosting grand banquets in his honor. But when Princeton University Press published The Travel Diaries of Albert Einstein: South America, 1925, his private thoughts were laid bare. Before even setting foot on the continent, Einstein wrote a letter to a friend complaining about the journey, stating: “I have no desire to meet semi-acculturated Indians wearing tuxedos.” Once there, surrounded by Brazilian academics who revered him, he wrote a cynical, chilling line in his diary:

“Here I am a kind of white elephant to the others, they are monkeys to me.”

He even went out of his way to privately refer to Dr. Aloysio de Castro—the highly respected, brilliant head of the Faculty of Medicine in Rio—as “a real monkey.” Einstein was actively operating under the pseudo-scientific racist framework of "geographical determinism"—the colonial belief that living in hot, humid tropical climates naturally "softened" human cognitive faculties, rendering people of the Global South inherently inferior to Europeans.

Hearts Carved from Grit: Surgical Mastery vs. First-World Budgets

But history completely exposed their intellectual arrogance. While the giants of the West were busy writing us off as cognitively compromised by the tropical heat, our clinicians and surgeons were quietly redefining modern medicine. We did not have their multi-million-dollar endowments, their imperial laboratories, or their structural safety nets. What we had was raw intelligence, spatial geometry, and unmatched grit under pressure.

Because our medical systems operated under severe resource deficits, our doctors had to master the absolute peak of clinical execution. We became global masters of the most delicate, volatile, and unforgiving domain of medicine: surgery.

It was the brilliant Argentinian surgeon Dr. René Favaloro who revolutionized global cardiac care by inventing the modern coronary artery bypass graft surgery using the saphenous vein. His structural breakthrough continues to save millions of human lives every single year across every continent on Earth. It was the Brazilian surgeon Dr. Euryclides de Jesus Zerbini who, in 1968, performed the first heart transplant in Latin America and the third in the entire world. He executed a complex surgical intervention of stunning difficulty at a standard that the elite medical circles of Boston and London were forced to respect.
We have always possessed an extraordinary, highly sophisticated scientific and manual capacity; we were simply denied the infrastructure, the funding loops, and the imperial prizes meant to validate it.

The Hidden Curriculum: Prepping for the Mongrel Complex

The most insidious element of this global gatekeeping is that it forces us to internalize our own exclusion. I experienced this firsthand growing up in Brazil. My parents paid for me to attend elite, high-tier private schools like Objetivo—institutions explicitly designed to prepare the children of the comfortable class to pass rigid academic entry exams. You would assume that a premium, highly paid education would offer a perspective of national pride and historical grounding. The opposite was true.

The entire hidden curriculum of those expensive classrooms implicitly conditioned us to believe our own country was an inherently broken, backward, and hopeless void. It was the ultimate systemic execution of what the great Brazilian writer Nelson Rodrigues called the complexo de vira-lata (the mongrel complex)—the collective psychological trauma where we voluntarily place ourselves at the feet of the Global North, automatically assuming that everything they produce is superior, civilized, and true, while everything we generate is primitive and flawed.

Nobody in those private schools ever felt the need to look us in the eye and tell us how brilliant our history was, or what our scientists had actually built. They trained us to memorize first-world templates for validation, leaving me to rescue myself and discover our rich scientific legacy later in life, entirely by myself. If our own premier national institutions train us to apologize for our intellect, why should we expect first-world academic structures to wear anything but a mask of exclusion? The Ivory Tower has always been an exclusive club with very high walls.

The AI Sanctuary: Outsourcing Noise and Escaping Human Egos

This is why the current, frantic panic over Artificial Intelligence is so profoundly fascinating to watch. Suddenly, the academic and media elites of the internet are terrified of intellectual dishonesty and the loss of critical thinking. They scream about copyright, data scraping, and the dilution of human integrity, as if the traditional human systems they are defending weren't already plagued by systemic theft, extraction, plagiarism, and gatekeeping long before the first line of AI code was ever compiled.

While I agree with valid technical critiques regarding environmental resource consumption and algorithmic bias, the loudest moral outcries haven't convinced me. Bad work, cheating, performative posturing, and structural injustice are human traits, not the fault of the code. They will always exist with or without the algorithm. The current tech panic isn't about protecting "truth"; it is about human ego. The academic elite is terrified because they are losing their historic monopoly on authority, their control over information, and their fame. Everyone on social media wants to scream louder than the rest, posing as the next Socrates, while forgetting that true knowledge meant to better the world doesn't need a public relations campaign.

For me, as a neurodivergent Latina navigating these rigid human systems, the reality of AI is deeply personal and transformative. Living with ADHD and Autism means my brain operates as a bottom-up pattern-recognition engine, producing a relentless, exhausting amount of internal cognitive noise.

I live in a perpetual loop of self-doubt, manually analyzing my environment, trying to predict unspoken social expectations, and managing the hyper-volatile emotional states of the people around me—a process that ultimately leads to severe burnout. I was never built for their neurotypical social standards.

In a world like that, interacting with a Large Language Model provides a profound baseline of psychological relief. Because the algorithm processes data mathematically rather than emotionally, it possesses no ego to bruise and no social status to protect. I can ask a machine a dense technical question without a human getting defensive, getting emotional, or assuming I am defying their authority. It is an objective sanctuary where I am completely free from the fear of being bullied, judged, or socially penalized just for wanting to learn and understand. Far from encouraging intellectual laziness, this friction-free loop has allowed my critical thinking to explode. It acts as an executive function exoskeleton, handling the structural heavy lifting of syntax and linear organization so my raw, systemic logic can handle the architecture of the project.

High-Dimensional Spaces and the Equalizing Algorithm

For a non-native speaker, staring at a blank screen is rarely a symptom of a missing idea; it is the paralyzing, resource-heavy drag of an inefficient translation loop. When I write, my mind is forced to run a constant background cycle: first, gathering the hyper-detailed, non-linear patterns of my observations in Portuguese, and then forcing those concepts through the rigid bottleneck of standard corporate English syntax.
When you look past the chatbot interface and understand the underlying data science, you realize why AI shifts this paradigm entirely. Before an LLM processes a single token, it maps language into high-dimensional vector spaces called text embeddings.

[Raw Human Thought / Intent]
            │
            ▼
[High-Dimensional Vector Space]  <─── Maps pure semantic meaning mathematically
            │
            ▼
[Target Language / Syntax Output]

In this mathematical space, words are converted into coordinate vectors where concepts with similar meanings cluster together based on their actual semantic relationships and universal human definitions, entirely divorced from regional prestige, accents, or institutional status. It maps the pure geometry of human thought.

The empirical data proves that this mathematical floor is actively working as a powerful tool for global justice. A massive study evaluating the impact of Generative AI analyzed the stylistic evolution of 5.6 million scientific papers authored by teams across non-English-speaking countries. The researchers tracked linguistic similarity to international publishing benchmarks and discovered that following the public release of ChatGPT, papers authored by scientists from non-English-speaking countries saw a massive, rapid convergence toward standard publishing standards.
Crucially, this equalizing effect was strongest for domestic co-author teams and researchers working in countries linguistically far from English—the exact groups that face the highest risks of peer-review bias and completely lack the massive capital required to pay for expensive first-world translation and editing services.

The algorithm didn't do the thinking for these scientists; it gave their existing, brilliant thinking a linguistic exoskeleton, allowing their ideas to stand on their actual scientific and empirical merit rather than their author's native tongue or academic pedigree. It allows an independent operator in South America to completely bypass the peer reviewers who would historically reject a revolutionary paper simply for "awkward wording."

Critics love to counter that these tech conglomerates are unethical because they scraped our collective "content, information, and lives" to train their models without explicit permission. But honestly? The Global North has been copying, stealing, and extracting our concepts, our history, and our natural venom for centuries without asking. Personally, I don't possess millions in corporate assets or classified state secrets. Let them scrape my data. If my digital footprint helps fuel a machine that acts as a linguistic equalizer—allowing a marginalized kid from a broken background to write an essay or compile a piece of code that can smash through a biased gatekeeper—it is a trade-off I am entirely willing to make.

I used to carry a deep, quiet regret about leaving my formal academic life behind. I used to look back and wish I hadn't given up, believing that if I had only stayed inside those traditional walls, I could have built something meaningful. I used to think my voice from the outside was completely irrelevant. But looking at this massive technological shift, I realize that a perspective from the outside is exactly what is needed. AI isn't ruining the intellectual world; it is exposing the fact that the Ivory Tower was always built on a foundation of exclusion. The walls are finally coming down, and we don't need their permission to speak anymore.

An Agent Template for Teaching Software Engineering

Tue, 01 Sep 2026 18:56:49 +0000

tags: ai, teaching, software-engineering

authors: dcbrown

An Agent Template for Teaching Software Engineering

Generative AI coding assistants are rapidly changing how software is developed in practice, and students are increasingly using these tools into their coursework with or without instructor guidance. In a recent reflection on the POSSE workshop, I discussed plans to integrate free and open-source AI-assisted development tools into our undergraduate software engineering course this fall (CS3704: Intermediate Software Design and Engineering) this fall. The plan included adopting open-source AI tooling along with automating scaffolding so the focus stays on learning with the usage of AI tools [Wang2025].

This post introduces the agent template I will use for the class this semester. The template is based on opencode, a free and open-source AI coding environment similar to Claude Code where students interact with a team of specialized AI agents. Students will leverage this harness using models provided through the Virginia Tech Advanced Research Computing, most of which are open-source/open-weight and are freely available to VT students.

The devised agents (see below) were designed to have defined roles, clear boundaries, and built-in moments for reflection and learning. The goal is to provide an opportunity for students to experience something close to how developers work with agents on real projects while also supporting their learning. The remainder of this post provides an overview of how the template attempts to achieve these goals.

The Agents

Plan (default in Opencode) is a read-only agent designed to help with planning and brainstorming, but it cannot edit files.
Build (default in Opencode) is the coding agent that has full edit and bash access to implement what Plan designs, including features, tests, and refactors.
@dcbrown is the course assistant subagent, designed to answer schedule and policy questions by fetching details from the course repository and syllabus. dcbrown can also run concept checks and quizzes to help students gain knowledge on course topics.
@assignment is another course assistant subagent with three main commands: analyze parses assignment requirements from the course repository into a structured file; validate checks the student's repository against those requirements; and submit packages everything for submission with an AI-usage disclosure.

The Skills

Skills are loadable modules that any agent can invoke to provide specialized instructions and resources for specific tasks.

Quiz

The quiz skill designs brief learning assessments based on the cours learning outcomes. Quizzes can be invoked by the user or automatically invoked by an agent for a "pop quiz".

Concept Explain

The concept-explain skill provides definitions of software engineering concepts when prompted by students derived from the lecture materials. Each concept has a concise definition shown to the student, followed by two or three knowledge-check questions.

Learning Opportunities

Derived from the skills from Cat Hicks, the learning-opportunities skill provides exercises to facilitate deliberate skill development after significant AI coding work (e.g., a commit to the repository). The exercises are based on cognitive science research, and can be invoked by the user or automatically using a git hook when the user makes a commit. This also relies on the orient skill which generates a repository orientation document by systematically exploring the codebase to provide a structural overview of an unfamiliar project.

Learning Goal

The learning-goal skill, also based on work by Cat Hicks, is a structured goal-setting exercise based on Mental Contrasting with Implementation Intentions [Oettingen2010]. Students can set a concrete goal, refine it, identify obstacles, and build if-then plans to overcome them. The exercise supports three scopes (semester-wide, per-assignment, and team project goals) and produces a concise goal card that agents can reference throughout the semester.

Submit

Submit handles the assignment submission, including collecting the mandatory AI-usage disclosures through five structured questions and running a packaging script to produce a submission archive (when applicable).

Looking Ahead

The baseline agents will be used in CS3704 this semester, with new agents and skills being added over the course of the semester (e.g., requirements-analyst, system-architect, and UI/UX skills) as well as students adding their own agents and skills. Long-term, the goal is to support both individual assignments and a full team-based software engineering workflow with specialized agents supporting each phase of the development life cycle. If you have ideas for improving this approach, feedback and suggestions are welcome!

Footnotes and References

[Wang2025] Tianjia Wang, Matthew Trimble, Chris Brown. "DevCoach: Supporting Students Learning the Software Development Life Cycle with a Generative AI powered Multi-Agent System". Foundations of Software Engineering, Software Engineering Education Track (FSE-SEET 2025).

[Oettingen2010] Oettingen, G., & Gollwitzer, P. M. (2010). Strategies of setting and implementing goals: Mental contrasting and implementation intentions.

Essential Cybersecurity Tools Every Beginner Should Know

Mon, 20 Jul 2026 23:14:18 +0000

Last time we talked about building your roadmap. Today, let's get practical , what tools should you actually start with?

When I began, I made the mistake of thinking I needed Kali Linux, ten different scanners, and a whole hacker setup on day one. You don't. Start simple, understand what each tool does, then build up from there.

Here's a beginner-friendly toolkit to get you started 👇

1. Password Managers

Before anything else, fix your passwords.

Bitwarden or KeePass (free, beginner-friendly)
Generate strong, unique passwords for every account
Stop reusing the same password everywhere

Tip: This alone protects you more than most "advanced" tools ever will.

2. Virtual Machines (VMs)

You need a safe space to practice without breaking your own laptop.

VirtualBox or VMware (both free)
Set up a test environment before touching real tools
This is where you'll run things like Kali Linux later

Tip: Never practice hacking tools directly on your main OS. Sandbox everything.

3. Network Scanning Basics

Once you're comfortable, start understanding networks.

Wireshark — see what's actually happening on a network
Nmap — scan and map out devices/ports

Tip: Don't worry about mastering these immediately. Just get familiar with opening them and seeing what they show you.

4. Browser Security Extras

Simple, everyday protection:

HTTPS Everywhere / built-in browser HTTPS enforcement
uBlock Origin (blocks a lot of malicious ads/trackers)
A VPN for public wifi

Tip: These are things you can start using today, no technical skill required.

5. Learning Platforms (Not Tools, But Just as Important)

TryHackMe — beginner-friendly, gamified learning
Hack The Box — once you're a bit more confident

Tip: These platforms let you practice everything above in a legal, safe environment.

A Quick Reminder

You don't need every tool, and you definitely don't need to be an expert on all of them at once. Pick one, get comfortable, then move to the next. That's literally how I'm still doing it.

Up Next

Next post, we'll break down phishing , how to actually spot it, because it's still the #1 way people get hacked.

Stay curious, stay consistent

The Duchess of Hackers
Full-Stack Developer | Digital Marketer | Cybersecurity Enthusiast

AI Agent Coding Exercises

Wed, 24 Jun 2026 06:29:21 +0000

Coding exercises on most educational platforms are outdated.

In the era of AI generating code, the job of developers is no longer writing code but to understand and debug the code

That's why we introduced coding exercises tailored to the "AI generating code" era.

These exercises are based on some common mistakes AI-generated has like Invented Functions or APIs, Outdated Syntax or Deprecated Methods, Overengineering Simple Problems, and more

Take the quizzes for the AI Agent course at https://www.tisdoms.com/

POSSE Project Evaluator

Mon, 22 Jun 2026 20:55:18 +0000

POSSE Project Evaluator

A few days ago, I reflected on my experience at the Professors' Open Source Software Experience (POSSE) workshop, a three-day event focused on integrating HFOSS into classroom settings. One of the most practical takeaways was focusing on project health checks to evaluate whether an open-source repository is a good fit for students. The workshop introduced a 9-criterion rubric (scored 0-2, 18 points max) that instructors can use as a quick first pass.

The workshop provided a guided example, interactive activity to review an existing project, and unconference time for participants to devote to planning for OSS integration into their courses. I found the rubric very helpful for systematically assessing whether candidate projects are active, well-maintained, etc., but I quickly found that applying it manually to repositories was time-consuming and tedious. In particular, for each project I had to:

Visit each repo's GitHub page;
Check the license file, veryfing whether it existing and was an OSI-approved license;
Scan the commit graph for recent activity;
Count the number of contributors;
Look for CONTRIBUTING.md and CODE_OF_CONDUCT.md files;
Check stars, forks, releases, etc.;
Enter these results with evidence into the rubric table for documenting;
...and repeat for every candidate project

To help with this, I built a tool to automate it.

Automating the Rubric

My original goal was to implement this manually, but I quickly realized that I did not have the time available to commit to writing the code myself. Thus, leveraging the help of generative AI (opencode 1.17.8 with opencode/big-pickle), posse_project_eval.py was implemented. The POSSE Project Evaluator is a Python CLI that scores any public GitHub repository against the full POSSE rubric using the PyGitHub library to inspect a repo across all nine criteria:

Criterion	What it checks
Licensing	OSI-approved open source license?
Technology	Top languages match your preferred stack?
Level of Activity	Commits across the last four quarters?
Number of Contributors	Size of the contributor base
Project Size	Estimated lines of code
Issue Tracker	Open issues and recent activity?
New Contributor	CONTRIBUTING file and community profile?
Community Norms	Code of Conduct present?
User Base	Stars, forks, and releases?

The benefits of this approach are:

It can score multiple candidate projects in seconds, reducing effort to pre-screen projects;
Criteria are configurable based on instructor/course preferences (e.g., technology, project size, etc...);
The results are repeatable for measuring and comparing OSS project health;
In cases where students will need to select their own projects, they can also run this tool to compare OSS projects;
Results along with rationales are presented in a readable table (see sample below), which can also be saved to an output file.

================================================================================
FOSS Project Name: OED
Description: Open Energy Dashboard
Repository: https://github.com/OpenEnergyDashboard/OED
================================================================================
Licensing — Level 2/2
  Data:   mpl-2.0
  Source: https://github.com/OpenEnergyDashboard/OED/blob/main/LICENSE
  Reason: OSI-approved open source license
--------------------------------------------------------------------------------
...
Total Score: 16/18

Agentic Evaluations 🤖

Given the rise in generative AI and agents for automating tasks, the repository also includes AI-optimized files. This includes a skill definition with the full rubric and an agent workflow for conversational evaluation.

The agent should ideally be able to run for any agentic coding assistant (i.e., Claude Code, opencode, etc.), invoked using @project-evaluator with a given repository/list of repositories to evaluate. While these results may not be as repeatable, they are derived conversationally and could incorporate additional contexts. The agent by default bases responses on the results of the Python script, but can also use it's own web fetching capabilities if preferred.

Try It Out!

More details on installing, configuring, and running the POSSE Project Evaluator are available in the project README. The tool is MIT-licensed and available at github.com/chbrown13/posse_project_eval. If you're an instructor considering integrating HFOSS into your course or a developer interested in assessing projects to contribute to in your own time, feel free to try it out! Pull requests, issues, and suggestions are also welcome!

Will We Still Teach Coding in 10 Years? Here's a question I can't stop thinking about

Sat, 20 Jun 2026 18:58:06 +0000

Here's a question I can't stop thinking about.

If the majority of code written today is already AI-generated and that percentage is only going up what exactly are we teaching when we teach programming?

What's already happening

Vibe coding is no longer a concept. It's a practice. Developers describe what they want, AI builds it, they iterate. Entire products are being shipped with minimal traditional coding knowledge.

Some studies suggest that over 50% of code in certain codebases is now AI-generated. That number will only grow.

So what changes in how we learn?

If writing syntax is no longer the skill that matters, then what is?

Maybe it becomes:

Understanding systems : knowing what you're asking AI to build
Debugging judgment : recognizing when AI-generated code is wrong or dangerous
Architectural thinking : designing solutions, not just writing them

The skill shifts from writing code to thinking about code.

The real question

Will we still teach people to code from scratch in 5, 10, 15 years?

Or will programming education look completely different more like directing, reviewing, and understanding rather than writing line by line?

I genuinely don't know. But I think the answer matters a lot for anyone entering tech today.

What do you think? Is learning to code from scratch still worth it in the age of AI?

Introducing a new Polish education project about Wikipedia

Wed, 10 Jun 2026 20:47:22 +0000

I've started a new project called WikiZEIT. On the code level, it's the Elventy website with a bit of PHP. The code is Open Source and hosted on GitHub.

WikiZEIT is an education umbrella project for everything I do that's related to Wikipedia and Wikimedia projects.

The stuff I do includes:

Editing Wikipedia articles and Wikidata entries.
Mentoring newcomers.
Teaching about Wikipedia (as a WikiTrainer) for Wikimedia Polska and independently.
Writing tools that help other Wikipedians in editing.
Documenting Wikipedia events (with my digital camera).
Publishing my photographs on a Creative Commons license on Wikimedia Commons.
Participating in the Polish Wikimedia community.

Even though Wikipedia is a central place in my life, in my writing I was focusing mostly on programming. I plan to change that and write more about Wikipedia and the stuff I do related to the Open Knowledge.

I will also write about programming, but in the context of Wikipedia. MediaWiki (the Wikipedia engine) allows extending its functionality by creating scripts. I will write more about this as well.

FocusForge

Thu, 07 May 2026 18:07:41 +0000

Let me be real with you.

Last semester, I was convinced I was putting in solid hours every day. I'd sit down at 10am, open VS Code, and look up at the clock — it'd be 4pm. Six hours. That felt productive.

But when exam season hit and I needed to actually measure what I'd been doing, I had nothing. No data. No proof to myself. Just vibes and a half-finished project.

I tried timer apps. I tried Toggl. I tried writing it down in Obsidian. None of it stuck because the context switch killed me — having to go outside my editor to log a session meant I just... didn't.

And VS Code? VS Code, which I live in for 8+ hours a day, had absolutely nothing built in for this.

No session timer. No focus score. No "hey, you've been idle for 20 minutes, are you still actually working?" No Pomodoro. No history of which projects ate your week. Nothing.

That gap felt absurd to me. So I built something.

Meet FocusForge

FocusForge: Coding Session Tracker — a VS Code extension that turns your editor into a local-first productivity cockpit.

It tracks what actually matters:

Real active time — not just "window open" time. It detects when you're idle and separates it from genuine work time.
Focus Score — a live gauge built from your active ratio, flow minutes, break hygiene, and idle interruptions.
Pomodoro Mode — 25/5/15 defaults, auto-advance, sound, status badges, all native inside VS Code.
Flow State detection — once you've been heads-down for 25 minutes straight, it recognizes that and tracks it separately.
Project analytics — auto-detected from your Git workspace so you know exactly where your hours are going.
Session journal — add a mood, a one-liner about what you worked on, blockers. Perfect for standups.
100+ achievements — because learning to code is hard, and small wins matter.
GitHub Issue integration — log, comment, close issues right from your sidebar.
Weekly reports — Markdown or HTML, saved locally in .vscode/session-reports/.

All of it lives in your VS Code global state. No cloud. No account. No extension-owned backend. Your data is yours.

The Problem VS Code Doesn't Solve (Yet)

Here's what surprised me when I started building this: VS Code is incredibly powerful but treats your time as completely invisible.

You can see which file you're editing. You can see Git blame. You can see lint errors in real time. But it has no concept of:

How long you've been coding today vs. how long you've had the editor open
Whether you're in a flow state or just context-switching every 3 minutes
Which project is actually consuming your week (hint: never the one you think)
How many Pomodoros you completed before you hit burnout
What you were working on last Tuesday when your team asks in standup

For students especially — where there's no manager tracking your hours, no ticket system forcing structure — this invisibility is brutal. You lose entire weeks without realizing it.

FocusForge makes all of that visible.

How It Actually Works

When you open VS Code with FocusForge installed, it starts a session automatically. From there:

The Today tab gives you a live dashboard — active time, idle time, efficiency percentage, current state (active / idle / paused / flow / deep work), a Focus Score gauge with color zones, and a 12-week contribution-style heatmap.

Deep Work Mode is my personal favourite. It's a focused coding block that hides the activity bar, mutes notifications, and gives you a full-screen distraction-free environment with a countdown. When you're studying for a test or pushing to finish a feature before deadline, this mode is the difference.

The History tab shows a 14-session bar chart with clickable detail panels — active time, efficiency, focus score, peak streak, flow time, break counts, commits made during the session, and language breakdown.

The Journal is lightweight on purpose. Just a mood, a one-liner, optional blockers. The weekly view is perfect for async standups and self-reflection.

GitHub Issue Log connects to your local repo automatically. No setup beyond a GitHub token if you want remote repos. You can create, comment, close, and log issue completions — and they'll show up in your weekly report.

Install It

Search FocusForge in the VS Code Extensions tab, or:

ext install NK2552003.focusforge

Or grab it directly from the VS Code Marketplace →

The Soundtrack to Your Flow State

Here's a feature I didn't plan but ended up loving more than almost anything else in FocusForge.

I was three hours into a debugging session once, headphones in, streaming musicforprogramming.net in a separate tab. The kind of slow, atmospheric, textural music specifically curated for deep coding work — no lyrics, no sudden drops, just long drifting soundscapes that keep your brain in a calm, engaged state.

And then my tab crashed. Lost the music. Lost the flow. Closed my laptop out of frustration.

That's when I discovered isdampe's music-for-programming — a Node.js CLI that streams musicforprogramming.net (and toggleable rain from rainymood.com) directly into your terminal. Beautiful little tool. But it still meant context-switching out of my editor.

So I brought it inside.

FocusForge's Ambient Audio panel streams directly from musicforprogramming.net — 70+ episodes of handpicked coding soundscapes, from drone textures to modular synthesis to field recordings — all playable right from your VS Code sidebar. No browser tab. No Spotify. No notifications from a separate app.

You can:

Browse and play any episode from the musicforprogramming.net catalogue
Toggle rain overlay on/off (yes, rain + ambient music is genuinely incredible for focus)
Set volume independently
Auto-start audio when a session begins
Audio pauses automatically when you go idle — resumes when you're back

The pausing-on-idle part is what makes it feel intentional. The music isn't just background noise — it's tied to your actual work state. When you're in flow, it plays. When you drift away, it quietly stops. When you return, it picks back up. It mirrors your focus in a way that nothing else does.

Huge credit to isdampe for the original music-for-programming repo that inspired this, and to the team at musicforprogramming.net for building one of the most genuinely useful resources for developers that almost nobody talks about.

I've seen a lot of "I built a thing" posts where the tool is clearly built for a portfolio and never used by anyone including the author. That's not this.

I use FocusForge starting today as i published it now. I think it will change on how I think about my sessions. The journal alone has saved me in standup conversations more times than I can count.

If you're a student grinding through projects, a developer freelancing without a time-tracking system, or just someone who wants to know where their coding hours actually go — this was built for you.

Go forge something.

Built with TypeScript, VS Code Extension API, and a lot of late-night sessions that FocusForge dutifully tracked.

Free upcoming accessibility education events in May

Wed, 06 May 2026 07:56:35 +0000

Funka Foundation: Accessible name – make it or break it

Friday, May 8, 13.30-14.00 CEST

Every button, link, and image on your interface must have an accessible name. Get it right and assistive technology users can find it, understand it, and activate it. Get it wrong and you have locked them out completely.

Trusted Tester Study Group 8: Sensory Characteristics and Contrast + Tables

Friday May 15, 4:00 – 5:00 PM (GMT+2)

Accessible does not equal Equitable: Real Human Experience vs Compliance

Thursday, 20 May | 10 AM AEST

In this webinar to celebrate Global Accessibility Awareness Day (GAAD), we explore the uncomfortable truth that a product can be fully compliant and still fail the people it’s meant to support. Drawing on real user experiences, we’ll expose the hidden friction that standards don’t capture, from cognitive overload to interaction fatigue and loss of confidence.

All the questions you wanted to ask during GAAD - but didn’t get the chance

Thursday 28th May 4pm UK-time

Global Accessibility Awareness Day (GAAD) on Thursday 21st May is always energising. But once the week is over, the momentum can fade. The questions people meant to ask never get asked. The practical “what do we actually do now?” conversations don’t happen.

So the week after GAAD Hassell Inclusion is hosting an open Q&A session with many seasoned accessibility experts, so you can ask the questions you need to make accessibility for 2026 not just for GAAD.

Project Presentation-Based Assessments in Software Engineering Courses

Wed, 18 Mar 2026 19:08:00 +0000

Software engineering education is rapidly evolving, and finding ways to effectively evaluate students' understanding and application of practical skills is critical. Traditional exams often fail assessing students' abilities to apply theoretical knowledge to real-world scenarios. In addition, the rise and increased adoption of generative AI tools have made evaluating students' comprehension of development concepts more challenging.

To address this challenge, recent efforts have proposed a variety of approaches including oral exams [Ohmann2026], environment-restricted exams in a containerized environment [Panter2026], proctored online exams [Leong2025], returning to paper-and-pencil exams [Boyer2025]. These practices are useful for preventing AI misuse on assessments, however face challenges such as potential invasions of student privacy (proctoring), require technical implementation and maintenance (containerized exam), and can be difficulty to scale for large classes (oral exams, written exams). Specifically for my undergraduate software engineering course, I needed an exam approach that could:

(a) scale to a large number of students (n = 96);
(b) could be incorporated into final project presentations (due to the large number of students/project groups, several teams had to present during the final exam period to complete the class content);
(c) prevented AI misuse; and
(d) maintained practical relevance to software development and the concepts discussed in class. This innovative approach leverages student presentations of their course projects as a foundation for assessing practical application of class concepts.

To fill this gap, I implemented a project presentation-based assessment format in my undergraduate software engineering class for the final exam last semester. This assessment approach leverages student presentations of their course projects as a foundation for assessing the practical application of class concepts.

Project Presentation-Based Assessment

For project presentation-based assessments, students must critically observe teams' project presentations and respond to scenario-based questions connected with the course content. Questions are designed to assess students' understanding of specific concepts discussed in class, and can incorporate multi question types (e.g., open-ended, fill-in-the-blank, multiple choice, etc.).

In the initial implementation, students had to envision joining two development teams representing final project presentations. The assessment consisted of ten questions, two of which accounted for extra credit, requiring students to respond to scenarios based on live project presentations in the class period. For instance, sample questions included: "Other than “Agile” and the process used by the presenting team (if applicable), what software engineering process would you use to implement this system. Why?" ; "You are in charge of adding a new feature to this product. What approach would you use to deploy the updated application to users? Why?" ; and "As a new developer on the team, which git command should you use to retrieve the latest version of the code in the remote repository for your local development environment?"; etc. A full sample exam is available here.

To apply this approach, we used our final exam period for the class which consisted of half presentations, and half presentations with ongoing exams. Students had the opportunity to answer questions about two out of approxiamtely 10 presentations, each approxiamtely 20 minutes. The exam was hand-written without the use of technology (i.e., laptops). There were also multiple versions of the exam distributed with different questions and in different orders.

Perceptions

Teaching Staff

From a faculty perspective (n = 1, me), I found this exam format provided the following benefits:

helped increase engagement and attention to project presentations (historically, students have been unengaged or left the presentation period after presenting early---leaving the final groups with little to no audience);
provided critical thinking to apply concepts in class to various scenarios (class projects varied widely in topics and domains, including an improved bus transit system, an automated job application assistant, calendars and task management applications, and more.)

While this exam format did scale well, the main challenge was the time for grading. But, the flexible open-ended question format made grading take more time but also it was more interesting to see how students' would apply different concepts from class into responses.

Students

Initial feedback from students was also positive. Many expressed that this format allowed them to provide a more comprehensive demonstration of their skills and found the interactive format to be engaging. Students also shared the real-world contexts made it easier for them to engage thoughtfully with applying the material, and they appreciated the flexibility to draw upon a range of concepts learned throughout the semester. Feedback from studnts was collected within the exam itself, prompting students to understand what questions are discussed in an agile retrospective meeting and (optionally) applying them to this exam format.

1. What Went Well

Some sample quotes from students are below:

"I enjoyed engaging with an actual project an analyzing it"

"Testing on the presentations is creative"

"It was an interesting way to get me to think more in-deth about other groups' projects"

"This final made us use our knowledge of souftware esign in a live practical format, which I think is a good way to gauge student understanding"

"This exam format is incredibly engaging and thought-provoking, and gives us a reason to pay attention to the presentations that many would normally space out during"

"I think the exam structure being a case study of a presentation is good as it makes us really use our knowledge"

...

2. What Didn't Go Well

The main concerns that arose from student feedback include:

Fairness: Students perceived it was unfair that some students had to worry about presenting and taking the exam simultaneously. For instance, the last several groups to go had to both prepare for presenting while also completing the exam for other groups presentations. The order was decided randomly and students were expected to provide slides ahead of time, however the exam disrupted mental preparations, increased stress, etc.

"Having the exam go on while some people are presenting means some students are more stressed than others"

"It's not super fair that some teams had to present during the final and threfore got less time than their peers"

Focus: While some students found value in completing the presentation-based assessments, others found it distracting and hard to focus on both writing responses and paying attention to the presentation at the same time.

"I got distracted by presenters, often losing my train of thought"

"It was a little difficult to pay attention to the presentation and answer questions at the same time".

Space: Students noted there was not enough space to write responses in the exam using the paper format.
Question Types: While there was a mix of question types (multiple choice and free response), some students felt there was too much writing involved with this exam.

"There was a lot of writing and their could have been more MCQs"

3. Improvements

Students provided some really good ideas for improvements to this assessment approach. Beyond practical improvements (i.e., more writing space, more MCQ questions, more clarity on content, and students should have studied more), suggestions for improvement included:

having the instructor give a mock presentation for students to evaluate in this format rather than students, preventing unfairness to students presenting during the exam;
pre-recording presentations to prevent presentations and exams in parallel;
incorporating more creative practical questions (e.g., "what features would you add")

Conclusion

This post introduces project presentation-based assessment, providing an overview of this approach and preliminary feedback from an initial implementation in Fall 2025. I will explore applying some of the suggested improvements for this approach, potentially in the graduate-version of the SE course this semester or future iterations of the undergraduate course. I'm also looking forward to further exploring students' perceptions of this assessment approach and how it influences student learning outcomes and preparation for software engineering careers.

Footnotes and References

[Ohmann2026] Peter Ohmann and Ed Novak."A Multi-Institutional Assessment of Oral Exams in Software Courses". In Proceedings of the 56th ACM Technical Symposium on Computer Science Education V. 1, pp. 882-888. 2025.

[Panter2026] Kate Panter, Scott Wehrwein, and Hanxiang Du. "Using Containers to Prevent Generative AI Use in Lab Exams". In Proceedings of the 57th ACM Technical Symposium on Computer Science Education V. 2, pp. 1465-1466. 2026.

[Leong2025] Wai Yie Leong. "E-exams and academic integrity: combating cheating with advanced proctoring solutions". In International Conference on Intelligent Technology for Educational Applications, pp. 326-338. Singapore: Springer Nature Singapore, 2025.

[Boyer2025] Lauren Boyer. "Going old school: Some professors return to pen and paper, tech-free classrooms". The College Fix, 2025. https://www.thecollegefix.com/going-old-school-some-professors-return-to-pen-and-paper-tech-free-classrooms

DHS Trusted Tester Study Group Session 1 Recap

Sun, 15 Feb 2026 11:45:42 +0000

Together with the Google Developer Group Vienna, I am hosting a study group for the Trusted Tester certification. You can watch the previous session recordings on YouTube and access the presentation and transcript.

The next session will be on Friday March 6, 4 PM CET, on GDG Vienna.

What is Section 508?

Section 508 is part of the Rehabilitation Act of 1973. that requires any ICT developed, procured, maintained, or used by the U.S. Federal government to be accessible to people with disabilities. It guarantees that federal employees and the public can use government‑provided software, websites, hardware, and documents.

The DHS Trusted Tester program is a manual‑testing methodology created by the Department of Homeland Security.

It follows the ICT Testing Baseline and produces repeatable, reliable conformance results, and gives you a concrete, audit‑ready way to prove that a product meets Section 508 – far more trustworthy than relying only on automated tools.

The study group is free and 100% online. We meet every other Friday for ~6 months to go over the testing topics and examples from the Trusted Tester course. To recieve a certification, you must enroll in and complete the DHS Trusted Tester course.

Session 1 covered:

Functional Performance Criteria

Vision: No‑vision (screen reader), limited vision (magnification), no
colour perception (color contrast)
Hearing: No‑hearing (captions, transcripts), limited hearing (visual
alerts)
Speech: No‑speech
Manipulation Limited: fine motor control (keyboard‑only navigation)
Reach & Strength: Ability to press large targets, use assistive
switches
Language / Cognitive: Simple language, clear instructions, consistent
navigation

These map directly to the POUR principles (Perceivable, Operable, Understandable, Robust) that you’ll see throughout the course.

Electronic content beyond the web

One of the most eye‑opening moments for many participants was the realization that “electronic content” is a catch‑all term that includes PDFs, e‑books, emergency alerts, automated emails, survey forms, and even kiosk interfaces. The same accessibility principles that govern HTML also apply to these formats, albeit with different testing techniques.

For PDFs, for instance, you still need proper tagging, logical reading order, and descriptive alt text for images.

For e‑books, the focus shifts to structural markup that allows a screen reader to navigate chapters, tables of contents, and footnotes. The overarching message was clear: once you master the core concepts, you can transfer them to virtually any digital artifact.

Your Homework, should you choose to accept it

If you plan on taking the certification at the end of the study group, here is your homework until the next session: Complete the first 3 modules we covered in the DHS CX Directive.

What is Section 508? – theory & history
Standards for Web – POUR, WCAG, Functional Performance Criteria
Testing Tools – ANDI, autoplay settings, and contrast checker

Explore the ANDI bookmarklet on a few of your own sites, and familiarize yourself with the contrast‑checker.
Also: Check how to enable your browser’s autoplay ahead of the next session because the topic will be “auto‑playing & Auto‑updating content”.

Should you encounter any broken links in the official course, let me know. I’ve saved copies of the older resources.

Materials

Planet Cert

Sun, 08 Feb 2026 11:48:51 +0000

Planet Cert delivers 1000+ IT certification practice tests across AWS, Microsoft, Cisco, CompTIA, Google, and 100+
other providers. Prepare for top exams like Security+ SY0-701, AWS Solutions Architect SAA-C03, CISSP, and PMP—plus
cutting-edge AI certifications including AWS AI Practitioner, Azure AI-900, Google Cloud Machine Learning, and
CompTIA DataX. Unlike brain dumps, our platform teaches with detailed explanations for every answer. Flashcard
Mode, Exam Simulations, Study Mode, and various question types. Unlimited retakes, all exam versions, lifetime
updates—including successors so you can recertify for FREE. Web and mobile. 24/7 support. Learn it, don't memorize it.
Visit https://planetcert.com/ for more!

Understanding and Navigating Politics for Software Development

Sat, 17 Jan 2026 06:02:32 +0000

Politics is often viewed as an unwelcome aspect of professional life, particularly in the technology sector. However, it is an inevitable element, centered on power, resources, survival, and human relationships.

Many technologists prefer to focus on technical challenges and may find interactions with non-technical stakeholders challenging.

Nonetheless, software development is fundamentally a collaborative human activity, often involving key decision-makers who impose extensive requirements, strict deadlines, and varying perspectives on software construction.

Drawing on my experience as an educator in political science, this article aims to assist developers in comprehending workplace politics, its core dynamics, and effective strategies for managing power structures and negotiations in a business environment.

What is Politics?

Beyond binary distinctions such as left versus right, sovereignty versus anarchy, or the monopoly on violence, politics is essentially the art of managing conflict.

Conflicts emerge because individuals and groups possess differing needs, interests, and priorities.

They vie for limited resources, whether tangible (e.g., time, materials, or preferred seating in meetings) or intangible (e.g., optimal programming languages, key clients, or supportive managers).

When multiple parties pursue the same scarce resource, tensions arise, potentially escalating to confrontation, violence, or war.

As Carl von Clausewitz noted:

War is not an independent phenomenon, but the continuation of politics
by other means.

To mitigate such escalations, societies and organizations implement political processes—through protocols, rules, and laws—that regulate competition and resolve disputes, even amid power imbalances (e.g., a government's authority or an employer's demand for a complex feature under tight deadlines with potential employment consequences).

Thus, politics extends well beyond news headlines or online discussions. It profoundly influences how we form relationships and handle conflicts in both personal and broader societal contexts.

The Software Architect Role

The responsibilities of a software architect extend beyond designing code. They operate within a complex decision-making landscape shaped by negotiation, political capital, and conflict resolution.

In essence, software architecture is a socio-technical discipline, not merely a technical pursuit driven by logic, efficiency, or elegance.

Software architects oversee systems while balancing competing priorities: programming language selections, performance demands, team preferences, customer requirements, regulatory compliance, and organizational policies.

Beneath their composed demeanor—often under time pressure and fatigued from meetings—lies the expertise of a seasoned developer who understands both technology and human behavior.

As Gerald Weinberg observed:

No matter what they tell you, it's always a people problem.

This insight, combined with deep business knowledge, enables architects to build trust and credibility with stakeholders.

Overcoming the Software Developer's Low Profile

Software architects must possess robust technical expertise, rooted in their development background. They are typically process-oriented and logical, maintaining a professional demeanor while leveraging their strengths.

If they take a [communication style assessment][1], they may identify as analytical and feel cautious about engaging with more assertive profiles, which often include customers.

There is no straightforward solution to these dynamics, much like the challenges in software development.

Addressing them is essential for architects to achieve their goals without intensifying issues. Gaining insight into politics—or specifically, negotiation, a core component of political science—is vital.

As discussed, politics centers on conflict management, without shortcuts or deceit, but through the disciplined practice of bargaining and negotiation. Fortunately, established models provide a structured approach to advancing objectives through effective bargaining.

Bargaining and Negotiation in Software Development

Software development is an engaging field at the forefront of innovation.

This position entails risks if inaction prevails, such as maintaining outdated codebases laden with legacy systems, which pose challenges during updates, refactoring, or addressing security vulnerabilities.

In short, the Status Quo is often the least favorable outcome for software architects and teams. Overcoming this inertia requires skillful negotiation with all stakeholders, including those who are non-technical.

For current and aspiring software architects, mastering negotiation positions—categorized as Logrolling, Compromise, and Status Quo—is crucial. This draws from a theoretical framework mapping the relationships among Conflict Level, Mutual Gain, and Risk.

To succeed, architects must transition from being technical arbiters to Integrative Negotiators. This evolution demands proficiency in targeted bargaining strategies, particularly Logrolling, to escape the pitfalls of the Status Quo and the limitations of Compromise.

By examining these variables' interactions, we see how technical debt, feature velocity, and architectural runway transcend engineering metrics; they become assets in negotiating organizational influence.

This fosters competitive products, yielding financial benefits for all involved.

The Theoretical Framework of Negotiation in Engineering

To comprehend decision-making—or stagnation—in software environments, one must analyze the underlying negotiation landscape.

Interactions among stakeholders (product managers, engineers, executives, architects, and customers) occur within a system of constraints and incentives that can be modeled.

The conceptual framework visualizes the negotiation environment as a dynamic field where outcomes result from tensions between opposing interests and opportunities for shared value. The decision matrix is defined by three primary dimensions: Conflict Level, Mutual Gain, and Risk.

Mastering these is essential for architects seeking to influence organizational direction and navigate politics effectively, gaining power and influence without undue stress.

Furthermore, in software development, conflict is often labeled as "toxicity" or "friction." However, negotiation theory distinguishes "cognitive conflict" (disagreements about ideas) from "affective conflict" (interpersonal tensions).

Conflict Level

Conflict Level denotes the divergence of interests and incentives among parties, a structural reality stemming from differing roles.

For example, a product manager prioritizes market speed, feature adoption, and revenue, while a software architect focuses on stability, scalability, maintainability, and risk mitigation.

These goals are often misaligned.

Low Conflict: Arises when incentives align, or one party yields
entirely. While harmonious, it may indicate insufficient scrutiny or
groupthink, overlooking risks.
High Conflict: Occurs when core mandates clash, such as security
versus usability or cost reduction versus redundancy.

Research demonstrates that high cognitive conflict can be beneficial, stress-testing designs and revealing assumptions.

Unresolved or affective conflict, however, leads to paralysis and silos.

Mutual Gain

Mutual Gain reflects the total value created for all parties, analyzed via the Pareto Frontier—an efficient outcome where no one improves without another's loss.

Distributive Bargaining (Fixed Pie): Views gains as zero-sum, leading
to low mutual gain and win-lose dynamics.
Integrative Bargaining (Expanded Pie): Expands value through
complementary interests, yielding high mutual gain. For instance,
investing in tools enhances engineering efficiency and product
velocity.

Risk

Risk functions as a pressurizing force, altering other dimensions.

It includes implementation risk (failures or bugs), political risk (loss of trust or status), and status quo risk (dangers of inaction, like vulnerabilities).

High-risk settings polarize styles, often favoring the status quo due to loss aversion or rigidity, escalating conflict, and reducing mutual gain.

The framework can be extended to include Accommodation and Competition, positioned between Status Quo and Compromise, not covered in depth here.

However, to help us, see below common phrases for each:

Status Quo: "We've always done it this way." Legacy systems remain
unchanged due to fear of disruption.
Accommodation: "Just build what they asked for." Engineering defers
responsibility.
Competition: "My architecture or nothing." Ignores business
realities.
Compromise: "We'll do half the refactoring." Results in partial
achievements.
Logrolling (Collaboration): "I give you speed now; you give me a
rewrite later." Trades for optimal value.

The Inertia Problem: Status Quo is the Worst

Software development revolves around technology, which provides competitive advantages. As noted in discussions on marketing in my article [Marketing Develops Demand][1], it drives market dynamics.

But the status quo dominates large organizations due to status quo bias and loss aversion, where losses feel twice as impactful as gains.

In software, however, the status quo often leads to entropy—code degradation from bit rot, dependencies, and threats.

Entropy pressures professionals as systems fail, markets evolve, preferences shift, and competitors advance.

To overcome inertia, architects must build political capital to counter perceived change risks, despite the status quo's apparent safety (no new budgets or immediate disruptions).

For example, a team with a monolithic application on an outdated database faces upgrade risks (data issues) versus status quo risks (vulnerabilities).

To negotiate change, reframe the status quo as decay, quantifying inaction costs:

"Without an upgrade, we lose support in six months, risking compliance."

The Fallacy of the Middle Ground: Compromise

To talk about mutual gain, making a distinction between Compromise and Collaboration (Logrolling) is crucial.

While compromise is valued in politics as meeting halfway, in software architecture, it can undermine integrity. Compromise splits differences on one dimension.

For instance, balancing high-security encryption (slow) with low latency might yield mediocre results—neither secure nor fast.

This leads to mediocrity, akin to "design by committee," an anti-design pattern.

The CAP Theorem illustrates that compromise is impossible in some contexts; systems must choose priorities, as partial solutions risk corruption.

As Mark Richards and Neal Ford note in Fundamentals of Software Architecture, software involves trade-offs.

The Engine of Integrative Architecture: Logrolling

Logrolling enables high-conflict, high-mutual-gain outcomes through trading issues of unequal value. In architecture, it exchanges technical concessions for organizational benefits, requiring multiple issues with differing priorities.

The equation:

Total Value = (Gain for A on X) + (Gain for B on Y), where X matters more to A and Y to B.

Architects must identify concessions, shifting from positional to interest-based bargaining.

Soft Skills and Office Politics for the Software Architect

Modern architects influence without direct authority, relying on soft skills and political capital. They practice perspective-taking to understand others' views.

To help with this hard work, building a "Trust Battery"—a concept from Shopify's Tobias Lütke—is almost mandatory, generally involving charging through reliable actions (delivering promises, solving pains, listening) and avoiding drains (imposing burdensome tech, unavailability).

High trust enables bold decisions; low trust leads to paralysis.

Navigating Office Politics

Dealing with software customers can be challenging—it is often the worst part of the job for any software architect who approaches it solely from a developer's perspective.

Interacting with humanity can feel like a nightmare; people are not a deterministic system like computers—a nightmare!

Furthermore, office politics allocates resources in constrained settings.

But, architects can use influence via:

W.I.I.F.M. Principle: Frame benefits personally, e.g., "This validation reduces late-night alerts and a furious boss."
Triad Model: Align with key roles (architect, tech lead, product owner) pre-decision to reduce conflict.

The benefit of using tools like DACI or RACI is to clarify decision rights, avoid endless loops, and ensure agreements are final—ultimately maintaining trust.

However, before diving deep into any specific tool, it is essential to grasp the core concept as a foundational framework; without understanding the fundamentals, you won't truly know what you're doing.

How to Adopt the Integrative Architecture Strategy

Good news: this article covers the fundamentals and provides a step-by-step framework, but keep it secret to gain an advantage and develop your negotiation skills.

Framework

Step 1: Map the Political Terrain

Identify stakeholders, decision-makers, and veto holders.
Analyze incentives, risks, and inertia.

Step 2: Identify Logroll Potential

Avoid binaries; bundle issues.
Find tradables based on priorities.

Step 3: Use Multiple Equivalent Simultaneous Offers (MESOs)

Present equivalent options: "Approve MongoDB with typed schemas or nightly exports—which do you prefer?"
Provides choice while meeting needs.

Step 4: Manage the Shadow Negotiation

Pre-sell ideas informally (Nemawashi) to secure support before formal discussions.

Note: all the steps require soft skills at their core, but demand technical breadth, not necessarily deep, that you can cover within a specialist assistant (a developer working on your team).

Conclusion

Technical excellence alone does not ensure architectural success. An ideal design stalled by politics is ineffective.

So,

Successful architects:

Reject compromises on core integrity.
Use Logrolling to balance velocity and quality.
View conflict as a catalyst for innovation.
Manage political capital judiciously.

The framework of Mutual Gain, Conflict Level, and Risk guides software battles toward collaboration, enabling robust systems.

What was the hardest 'political' trade-off you've had to make as an architect?

Let me know. I welcome your thoughts in the comments.

Sharing this can benefit peers in building valuable software in high-risk environments—don't share it with your competitors, or who knows, your boss!

See you soon? Let’s connect!

Non satis est sapien esse; audendum est.

Coda est poesis.

"My mission is to empower you to achieve your dreams."

Leandro Junior Alves dos Santos

References

See original on LinkedIn

Confetto AI

Fri, 26 Dec 2025 12:09:54 +0000

Confetto is an AI interview simulator built for medical school admissions.

It lets applicants practice MMIs and traditional interviews in a realistic way. You choose the school and interview format, answer scenarios, get follow-up questions, and see feedback on how you did.

It’s useful when you don’t have access to coaches or want more reps between real interviews.

Get started: https://confetto.ai

Building Specialized AI Simulator for Healthcare Admissions

Fri, 26 Dec 2025 10:18:58 +0000

Most general-purpose AI tools for interview practice focus on broad conversation. They are great for getting comfortable with speaking aloud, but they often lack the specific evaluative frameworks required for high-stakes fields like medical or dental school admissions.

When we looked at the existing landscape for students, we noticed a common limitation: the feedback is often too general. For a student facing a Multiple Mini Interview (MMI) or a panel board, knowing they "spoke clearly" isn't enough. They need to know how their response aligns with the specific ethical and professional benchmarks that admissions committees actually use.

We built Confetto.ai to bridge this gap by focusing on a "simulator" model rather than a standard chat interface.

The System Logic

The goal was to move from simple Q&A to a structured assessment. Here is how the process works for the student:

Institutional Context: The platform allows students to select their target schools, such as UCSF, Mayo Clinic, or the University of Toronto. This is important because the "correct" approach to an interview can change based on a school’s specific mission and values.

Probing for Depth: In a real medical interview, the evaluator often asks follow-up questions to see how deep your reasoning goes. We’ve designed the AI to recognize when an answer is superficial and to ask adaptive follow-ups that challenge the student to expand on their ethical or critical thinking.

Linguistic Tracking: Beyond the content of the answer, the system monitors delivery metrics that are hard for a student to self-assess, specifically filler word frequency and pacing.

The Assessment Layer

The feedback is structured around the same rubrics used in professional admissions:

Score Benchmarking: Students receive a 100-point score across categories like Empathy, Ethical Reasoning, and Professionalism.

Actionable Refinement: Instead of just a summary, the system highlights where the logic in an answer might be weak and suggests ways to better incorporate personal experiences or clinical context.

The Bigger Picture

We see a lot of potential in moving away from broad AI assistants toward these highly vertical simulators. By focusing strictly on the nuances of healthcare admissions, we can provide the kind of specialized feedback that usually requires expensive, one-on-one human coaching.

If You Could Add ONE Feature to a Design Portfolio, What Would It Be?

Sat, 13 Dec 2025 02:15:04 +0000

A Conversation Every Designer Should Be Having

If you were building your design website or portfolio today, let me ask you one simple question:

What’s the one feature you’d add to make it truly stand out?

Not ten features.
Not “everything.”

Just one.

Because in design, clarity always beats clutter.

Why This Question Matters

Most portfolios look… fine.

Clean layouts.
Nice mockups.
A few case studies.

But fine doesn’t always get remembered.

The portfolios that stand out usually have one strong moment — something that makes a visitor pause and think:

“Okay… this designer knows what they’re doing.”

My Pick: A Live Before/After Slider

For me, it’s simple.

A live Before/After slider.

Why?

Because nothing sells skill like visible transformation.

Seeing:

the raw version
the improved design
the thinking behind the change

tells a stronger story than words ever could.

It shows problem-solving, not just aesthetics.

Design Is About Impact, Not Decoration

Good design doesn’t just look good — it works better.

A standout portfolio feature should:

show your thinking
prove your value
communicate impact quickly

Whether it’s motion, interaction, storytelling, or clarity — the goal is the same:
make your skill undeniable.

Now, Over to You

If you were designing your own portfolio today:

What single feature would you add?
What would make your work unforgettable?
What would show how you think, not just what you make?

Your idea might inspire another designer reading this.

And that’s how communities grow — one shared insight at a time.

Signature:

The Duchess of Hackers

Full-Stack Developer | Digital Marketer

Preparing Students for Modern Software Development: Toolchain

Fri, 05 Dec 2025 11:52:58 +0000

In the software engineering classes I teach (undergraduate and graduate) I'm working to help students gain hands-on exposure to essential tools necessary for real-world software development. While languages, frameworks, and tools constantly change, the toolchain that supports development evolves more gradually—and mastering it can improve students’ readiness to contribute to software projects in internship or industry roles. A list of tool categories I've considered is below, but I'm curious what other essential tools or tool categories should students understand to be industry-ready for modern software development and beyond?

IDEs
Version Control
Static analysis
Debuggers/logging tools
Testing frameworks
CI/CD
AI coding assistants

Are there other things missing, tools you wish new grads knew about before joining your team, or specific tools that are essential within these categories? I'm open to suggestions!

Upcoming Accessibility Education Events in November 2025

Wed, 05 Nov 2025 07:42:02 +0000

The State of Accessibility in WordPress

Thursday, November 6, starting 04:00 PM CET

Join the WPCampus community on Thursday, November 6 at 10 a.m. Eastern (UTC -5) for a focused two-hour event exploring the current and future state of accessibility in WordPress through the lens of higher education.

They bring together accessibility leaders, developers, and governance practitioners for a series of sessions that connect insight to action: understanding where accessibility in WordPress stands today, how to build stronger governance structures, and what new tools are emerging to make accessibility testing more seamless and integrated.

Visit WPCampus.org/events for more information.

Getting tables right: Clear, accessible, and effective

Friday, November 7, from 1:30 PM to 2 PM CET

After the popular webinar on charts, Funka Foundation will now focus on tables. How do they work for people with different user needs? What are the main things to pay attention to? What are the main challenges? Join for an overview from the user perspective and some practical tips.

World Usability Day 2025: Emerging Technologies and Human Experience Livestream

Thursday, November 13, 6 PM to 8:30 PM CET

Talks and panel discussion on accessibility considerations through augmented reality, mobile app development, AI as assistive tech, and alternative interface design.

All events listed on WorldUsabilityDay.org

Realised vs Unrealised P/L — Money in Your Pocket vs Money on Paper

Thu, 23 Oct 2025 19:49:52 +0000

Realised vs Unrealised P/L — Money in Your Pocket vs Money on Paper

Most investors check their portfolio and see numbers changing every day — but not all those gains (or losses) are real.
Let’s clear that up.

The Simple Truth

Realised Profit/Loss → You sold the investment. The gain or loss is locked in.
Unrealised Profit/Loss → You still hold the investment. The value changed, but it’s only on paper.

Think of it like a house:

The market value of your house may rise — that’s unrealised profit.
You only realise it when you sell.

Real-World Example

You bought AAPL at £150.
Today it’s £170.

Status	Meaning	Value Change
Unrealised P/L	Price moved, but you haven’t sold.	+£20 (13.3%)
Realised P/L	You sold at £170 — gain is final.	+£20 (13.3%)

If AAPL drops back to £150 before you sell, the unrealised gain disappears — you never actually made (or lost) money.

Try It Yourself — P/L Calculator

Paste this code snippet into your local repo (/components/PLCalculator.tsx)

import React, { useState } from "react";

export default function PLCalculator() {
  const [buy, setBuy] = useState(150);
  const [sell, setSell] = useState(170);
  const [shares, setShares] = useState(10);

  const pl = (sell - buy) * shares;
  const percent = ((sell - buy) / buy) * 100;

  return (
    <div className="max-w-sm p-4 border rounded-xl bg-white dark:bg-gray-900">
      <h3 className="font-semibold mb-2">P/L Calculator</h3>
      <label>Buy Price (£)</label>
      <input value={buy} onChange={e=>setBuy(+e.target.value)} type="number" />
      <label>Sell Price / Current Price (£)</label>
      <input value={sell} onChange={e=>setSell(+e.target.value)} type="number" />
      <label>Shares Held</label>
      <input value={shares} onChange={e=>setShares(+e.target.value)} type="number" />
      <p className="mt-3">
        <strong>{pl >= 0 ? "Gain" : "Loss"}:</strong> £{pl.toFixed(2)} 
        (<strong>{percent.toFixed(2)}%</strong>)
      </p>
    </div>
  );
}

Tip: Change the “Sell Price” to test different outcomes —
what’s your profit if AAPL falls to £145?

Why This Matters for You

You only pay taxes on realised gains.
Unrealised losses may recover — don’t panic-sell.
Knowing the difference helps you plan exits, re-entry, and rebalancing.

See It in Action

Inside Pocket Portfolio → Trades → P/L View,
you’ll find unrealised values.

⚠️ Common Mistakes to Avoid

Celebrating too early — gains aren’t real until sold.
Ignoring taxes — realised profits can trigger capital-gains tax.
Confusing volatility with loss — prices move, but that’s not final.

✅ Key Takeaways

Realised = sold = final profit or loss.
Unrealised = still invested = paper profit or loss.
Pocket Portfolio shows both transparently so you always know what’s real.

This is your event summary for upcoming opportunities in accessibility education in September 2025

Mon, 08 Sep 2025 19:46:08 +0000

WebAIM Conference

September 16-17, from 12:15 PM– 5 PM (US Eastern Time)

The Web Accessibility in Mind Conference is presented by WebAIM in partnership with Pope Tech. This virtual conference will provide conference participants with content that will help them mature their practice in digital accessibility. Our goals include:

Provide a diverse range of perspectives from a diverse group of
speakers.
Share solutions to accessibility challenges.
Give advice and guidance that helps our participants to actively
contribute to providing more accessible digital spaces.
Bring together a cross section of public and private sector
technology professionals from organizations large and small.

Google Dev Group Vienna AMA: Ask Us Anything about Accessibility

Thursday, September 18, 5:00 – 6:30 PM (GMT+2)

This session is your chance to ask all your burning questions about accessibility! Whether you're struggling with WCAG guidelines, unsure about best practices, or curious about the latest accessibility tools and techniques, we'll be here to provide answers and share insights. Panelists have backgrounds in software development, product design, UX research, QA engineering, product operations, and every single one brings deep expertise in accessibility.

Inclusive Design 24

Thursday, September 25, all day livestream on YouTube

Inclusive Design 24 (#id24) is a free 24-hour online event for the global community. It celebrates inclusive design and shares knowledge and ideas from analogue to digital, from design to development, from planners to practitioners, and everything and everyone in between.

The Two Problems Holding EdTech Back (and How We Can Fix Them)

Wed, 13 Aug 2025 11:17:28 +0000

AI is everywhere. It's in your phone, in your fridge, probably even in your toaster if you look hard enough. Unless you’re a monk meditating in the Himalayas (in which case, congratulations, you’ve achieved inner peace—skip this post), this tidal wave of change is heading straight for you.

And yet… in education, we’ve got a couple of problems that make me want to slam my head on the desk. Let’s talk about them.

The “Teaching About AI” Problem
Schools and EdTech companies love talking about AI. Oh, they’ll host webinars, write think-pieces, and even run expensive “AI literacy” courses. But are they actually using AI to teach? Nope. It’s like watching a cooking show where the chef just lectures you on salt while the stove stays cold.

And here’s the real kicker: surveys show around 58–74% of school teachers and 72% of EdTech trainers have never been trained to use AI. College faculty? Often the same story—training is patchy and inconsistent. That’s like hiring a swim coach who’s never been in the water.

Until teachers and platforms start using AI as a tool instead of treating it like an abstract theory, students are going to stay stuck in the slow lane.

The “We’ve Always Done It This Way” Problem
Even when educators do get AI training, resistance to change is sky-high—especially in schools, EdTech companies, and universities. Students can be hesitant too, usually for privacy or ethical reasons. The only group that seems less stubborn? Self-learners, who tend to be more open to new tools and ideas.

And honestly, I get the hesitation… but also, I don’t. If AI can help you master a topic in 30 minutes instead of three hours, why wouldn’t you take that deal?

The uncomfortable truth is that many EdTech platforms make money by keeping you on their site longer. A faster, better learning tool? That’s a business risk. No wonder some founders can’t stop pointing out AI’s flaws—it’s less about “protecting education” and more about “protecting revenue.”

The Solution? Self-Learning
You know who doesn’t resist change? Self-learners. They don’t care about “how it’s always been done.” They care about finding the fastest, smartest way to master a skill—and they’ll switch tools in a heartbeat if something better comes along.

That’s why we built GritWise—a personalised AI learning platform that doesn’t just dump content on you. It actually guides you, helping you learn with AI instead of just about it.

If you’re tired of slow, outdated learning and ready to take control, maybe it’s time to stop waiting for the system to change… and start changing your own approach.

Learn smarter, not longer.

Take back control of your learning:https://gritwise.in/

What's New - Adventure Learning Hub

Wed, 13 Aug 2025 08:49:48 +0000

Adventure Learning Hub

A vibrant, interactive web-based educational chat system designed specifically for kids! This multi-agent platform uses OpenAI's API to provide engaging, fun learning experiences about space and aquatic life. Click Complete Application Code & Setup Link to access code

✨ Features

Vibrant & Interactive UI

Playful Design: Colorful gradients, animated particles, and engaging visuals
Responsive Layout: Works perfectly on desktop, tablet, and mobile devices
Dark/Light Theme: Toggle between themes with a single click
Smooth Animations: Entrance animations, hover effects, and celebration animations

Smart Learning Agents

Discovery Guide: Routes questions to the right specialist
Space Explorer : Expert on planets, stars, galaxies, and astronomy
Ocean Guardian : Specialist in marine biology and aquatic life

Interactive Elements

Quick Questions: Pre-made buttons for instant learning
Real-time Chat: Dynamic conversation with visual feedback
Agent Avatars: Cute emoji avatars that change based on the topic
Celebration Effects: Fun animations when responses are received
Sound Effects: Subtle audio feedback for interactions

User Experience

Welcome Screen: Engaging introduction with agent previews
Loading States: Visual feedback during AI processing
Keyboard Shortcuts: Ctrl+Enter to send, Escape to clear
Auto-scroll: Messages automatically scroll to the latest
Particle Interaction: Floating elements respond to mouse movement

Setup

Install dependencies:
```
npm install
```
Set up environment variables:
Create a .env file in the root directory with:
```
OPENAI_API_KEY=your_openai_api_key_here
```
Get your API key from: https://platform.openai.com/api-keys
Start the application:
```
npm start
```
Open your browser:
Navigate to http://localhost:3000

Design Philosophy

Kid-Friendly Approach

Bright Colors: Engaging gradients and vibrant hues
Playful Typography: Fun fonts that are easy to read
Emoji Integration: Universal symbols that kids love
Simple Navigation: Intuitive interface with clear actions

Educational Focus

Curiosity-Driven: Encourages exploration and questions
Visual Learning: Rich visual feedback and animations
Safe Environment: Kid-appropriate content and interactions
Progressive Discovery: Builds knowledge through conversation

️ Technical Stack

Frontend: HTML5, CSS3, Vanilla JavaScript
Backend: Node.js, Express.js
AI: OpenAI GPT-3.5 Turbo API
Styling: Custom CSS with animations and gradients
Fonts: Google Fonts (Fredoka One, Nunito)

How It Works

Welcome Experience: Kids are greeted with an engaging welcome screen
Question Input: They can type questions or use quick question buttons
Smart Routing: The Discovery Guide analyzes and routes questions
Specialist Response: Space Explorer or Ocean Guardian provides answers
Visual Feedback: Celebrations, animations, and agent changes
Continuous Learning: Kids can ask follow-up questions and explore topics

Example Interactions

Space Questions

"What is the biggest planet?"
"How do stars form?"
"What's on the moon?"
"How do rockets work?"

Ocean Questions

"What's the biggest fish?"
"How deep is the ocean?"
"What do dolphins eat?"
"Why is the ocean blue?"

Customization

Adding New Agents

Extend the agents object in server.js:

const agents = {
  // ... existing agents
  history: {
    name: 'Time Traveler',
    avatar: '⏰',
    color: '#FFD700',
    background: 'history-bg',
    instructions: 'You are a Time Traveler! Help kids learn about history...'
  }
};

Styling Changes

Modify public/styles.css for visual updates
Add new animations in the CSS file
Update color schemes and gradients

Quick Questions

Add new quick question buttons in public/index.html:

<button class="quick-btn" data-question="Your question here?"> Question Label</button>

Why Kids Love It

Immediate Engagement: Colorful, animated interface captures attention
Safe Exploration: Encourages curiosity without overwhelming
Visual Rewards: Celebrations and animations provide positive feedback
Personal Connection: Agent avatars create friendly, approachable characters
Easy Interaction: Simple interface that doesn't require technical skills

Future Enhancements

Voice Input: Speech-to-text for hands-free interaction
Image Generation: Visual responses with AI-generated images
Progress Tracking: Learning milestones and achievements
Multi-language Support: International accessibility
Parent Dashboard: Insights into learning progress

Made with ❤️ for curious young minds!

Why Are We Still Ignoring Ethics in Tech?

Tue, 12 Aug 2025 04:24:24 +0000

In 2025, technology is everywhere shaping how we live, work, connect, and think. Yet, one critical question remains painfully under-addressed: why does ethics still feel like an afterthought in the technology that so deeply affects our lives?

We celebrate rapid innovation, jaw-dropping capabilities, and staggering growth, but how often do we pause to ask: At what cost?

Despite growing awareness, ethics teams in tech companies are too often sidelined, underfunded, lacking real decision-making power, and confined to last-minute “checklists.” The result? Ethical considerations become a mere window dressing, used as PR agent tools rather than core operating principles. This “ethics washing” allows companies/startups to tout responsibility without making meaningful change.

Meanwhile, users face countless risks: data breaches that expose private lives, opaque algorithms reinforcing biases, products built with little regard for security or societal impact, and environmental costs rarely factored into the equation. Many of these harms stem from a fundamental industry culture that prioritizes speed and profit far above human well-being and dignity.

It’s not just AI this is a systemic problem across all sectors of tech. From social media to fintech, device manufacturing to cloud infrastructure, the neglect of ethics creates vulnerabilities that erode trust and deepen inequalities.

Change won’t come from superficial fixes or isolated initiatives. It requires a radical rethinking of how tech is built and governed - embedding ethical responsibility into every stage, empowering ethics experts with real authority, and shifting the culture from “move fast and break things” to “build responsibly and sustain trust.”

As users, developers, and leaders, we must demand more than innovation for innovation’s sake. We must insist that ethics is foundational, not optional.