AI6 min read
The model is not the product
Everyone argues about which model is smartest. Almost nobody ships. The ninety things between a prompt and a feature.
Blog
What I think about when I build
Notes on software, AI, products, open source, and the Pharmacy × Tech life — studied at UNN, written in public by Nzubechukwu Cyprian.
Start here
Product building6 min read
Shipping is a skill, and I practised it badly for years
Finishing is a practice with techniques — the ones that turned my graveyard of half-projects into shipped products.
Pharmacy × Tech6 min read
Pharmacy by school, software by conviction
The post that explains who I am — why Pharmacy and software are one strategy, not two competing lanes.
Software engineering4 min read
The bug that taught me to read error messages properly
The three-line error that sat in my logs for a week, because I never read past the first line.
Open source5 min read
Cloning a repo is not a portfolio piece
A repository is evidence, not ownership. This is the rule the whole website is built on.
Pharmacy × Tech6 min read
Learning pharmacology made me better at AI systems
Dose-response curves, half-lives, and drug interactions are surprisingly good mental models for AI in production.
Latest
The model is not the product
Everyone debates which model is smartest. Almost nobody talks about the ninety other things standing between a prompt and a shipped feature.
AI6 min read
Read the postAll posts
Software engineering5 min read
Naming things is the whole job, sometimes
Most codebases don't have a naming problem — they have a thinking problem that shows up in their identifiers.
Product building6 min read
Shipping is a skill, and I practised it badly for years
Finishing things is not a personality trait — it's a practice with techniques. The ones that turned my graveyard of half-projects into shipped products.
Open source5 min read
Your first open source contribution doesn't need code
Docs, reproduction steps, and tested bug reports are contributions. The barrier you imagine is lower than the one that exists.
Pharmacy × Tech6 min read
Pharmacy by school, software by conviction
I didn't choose between the lab and the keyboard. The combination isn't a compromise — it's a strategy neither discipline alone gives you.
AI5 min read
My AI workflow is boring, and that's the point
No autonomous agents rewriting my codebase. No prompt wizards. Just a disciplined loop between a draft, a review, and a test suite.
Software engineering5 min read
Your README is your first interview
Before anyone judges your code, they judge your README. What a good one actually contains, from someone who reads a lot of them.
Product building5 min read
Nobody told me product decisions are reversible or not
The single question that unblocked my decision paralysis: is this door one-way or two-way? Everything else is noise.
Open source5 min read
Reading the issue tracker: the skill nobody teaches
Issue trackers are where projects tell the truth. Learning to read them well reveals what to contribute and what to avoid.
Pharmacy × Tech5 min read
What dispensing prescriptions taught me about validation
A prescription is an API call with life-or-death validation requirements. Pharmacy school taught me forms before code did.
AI6 min read
Prompting is requirements gathering with worse vocabulary
The teams that get good results from AI aren't better at talking to models. They're better at saying what they want.
Software engineering4 min read
The bug that taught me to read error messages properly
A three-line error sat in my logs for a week. It told me exactly what was wrong. I just never read past the first line.
Product building5 min read
The pricing page taught me more about users than analytics
Data told me what people clicked. The pricing page told me what they believed. Convincing beats tracking.
Open source6 min read
The anatomy of a pull request that gets merged
Small scope, clean description, tests, and the maintainer's style. What separates merged PRs from the ones that die open.
Pharmacy × Tech5 min read
Two vocabularies: clinicians and compilers
Being bilingual in health and tech means seeing the same problem described twice — and knowing both descriptions are partial.
AI7 min read
I gave an AI agent shell access. Here's what I learned.
A week of running an autonomous coding agent on a real project taught me where the ceiling is — and it isn't intelligence.
Software engineering5 min read
CSS taught me more about engineering than I expected
Specificity, inheritance, and the cascade are systems thinking in disguise. Understanding CSS made my other code better.
Product building5 min read
Every product I've shipped taught me to delete more
My best products have fewer features than my first drafts. Subtraction is the quality signal nobody markets.
Open source5 min read
Cloning a repo is not a portfolio piece
A local copy of a repository proves you downloaded it. What actually counts as evidence of your work — and how to present it.
Pharmacy × Tech5 min read
The stoichiometry of technical debt
Chemistry taught me that unbalanced equations explode. Unbalanced architecture does too — just slower, and with more meetings.
AI7 min read
RAG is a data-cleaning problem wearing a trench coat
Most retrieval-augmented systems fail because of chunking, duplication, and stale documents — not because of embeddings.
Software engineering5 min read
How I review my own code (and why you should too)
The cheapest code review available: you, twenty minutes later, with a checklist and no ego. It catches more than you'd think.
Product building6 min read
Building for Nigerian users: what tutorials don't cover
Data costs, payment rails, offline moments, and trust signals. The local realities that reshape every product decision.
Open source5 min read
What maintaining your own small library teaches you
Nobody uses it? Perfect. Publishing and maintaining even a tiny package teaches the responsibilities that side projects never do.
Pharmacy × Tech5 min read
Why health data needs engineers who understand clinics
A dose is not a number and a patient is not a row. The case for developers who've stood on the clinical side of the counter.
Software engineering5 min read
Version control is a time machine, not a filing cabinet
Commits that tell a story, branches that hold experiments, and the confidence to delete things. Git rewards the curious.
AI6 min read
What AI can't do for a solo developer
AI writes code fast. It still can't decide what's worth building, notice the bug you haven't hit yet, or care about your users.
Product building5 min read
The MVP is not a bad product — it's a question
Minimum viable doesn't mean minimum quality. It means the smallest thing that can generate a real answer about real demand.
Open source4 min read
The etiquette of saying no in open source
Maintainers say no all the time — politely, clearly, with reasons. Contributors need the same skill: how to accept it and how to do it.
Pharmacy × Tech5 min read
Exams and deadlines: how I schedule two careers
Semesters have seasons. Products have seasons. The calendar system that keeps both moving without either burning down.
Software engineering4 min read
The environment variable that cost me a day
It worked on my machine, failed in production, and the difference was one missing string. Now I fail fast on purpose.
Product building6 min read
How I do user research with no budget and no team
Five conversations beat five hundred survey responses. A solo developer's guide to hearing the truth without a research department.
AI5 min read
Evaluating AI output like a code reviewer, not a fan
A practical checklist for reviewing machine-written code: correctness first, then edge cases, then the quiet sins.
Open source5 min read
How I find projects worth contributing to
The best project to contribute to is one you use, that's healthy, and that needs what you can give. A practical filter.
Pharmacy × Tech5 min read
The placebo effect of good UX
Expectation shapes outcome in medicine and interfaces alike. What pharmacy taught me about designing for trust, not just function.
Software engineering6 min read
Databases: the four mistakes I made so you don't have to
No indexes, no transactions, timestamps without timezones, and deleting without a backup. A field report.
Product building5 min read
What I learned maintaining a product nobody pays for yet
Free products have users with opinions and none of the patience. The maintenance lessons that apply to everything I'll ever charge for.
Open source5 min read
Writing documentation people actually use
Good docs answer the question the reader has *now*. Task-first, honest about limits, and tested like code.
AI6 min read
Building with AI on a slow connection
AI-assisted development from Nigeria means timeouts, retries, and offline-first habits. Here's the setup that survives bad networks.
Pharmacy × Tech6 min read
Learning pharmacology made me better at AI systems
Dose-response curves, half-lives, and drug interactions are surprisingly good mental models for how AI features behave in production.
Software engineering6 min read
Learning a new framework in 48 hours
The method I use when a project demands a stack I don't know yet: transfer the concepts, skim the docs, build the smallest thing.
Product building6 min read
The launch that flopped taught me the most
I launched to silence: no crash, no users, no feedback. Diagnosing that silence was the most useful product education I've had.
Open source5 min read
Licensing for people who publish things
MIT, GPL, Apache — what the choice actually decides, and why shipping without a licence is a decision you already made badly.
AI6 min read
AI is changing how beginners learn to code — badly
The path from confusion to competence used to run through being stuck. AI removes the being stuck, and with it the learning.
Pharmacy × Tech6 min read
Building tools for the health workers around me
The clinics near campus run on paper and WhatsApp. What I've learned watching real health workflows strain against their tools.
Software engineering5 min read
Write code like the next reader is exhausted
The person debugging your code at 2am will be tired, under pressure, and possibly you. Optimise for that reader.
Product building5 min read
Features die quietly: recognising when to sunset something
Features don't announce their own death. The signals I watch for, and the discipline of retiring things with respect.
Open source5 min read
The contribution that taught me to write smaller patches
A 400-line PR taught me more about contributing than ten merged ones — by taking three weeks to die in review.
Pharmacy × Tech5 min read
The version control habit I learned from patient records
Health records taught me that history is sacred. Audit trails in medicine are older than git — and stricter.
AI5 min read
The quiet rise of small, boring, useful AI
The most valuable AI systems I use aren't impressive. They classify, extract, and route — and they save hours a week.
Product building4 min read
Why I write the landing page before the product
One page of honest writing is the cheapest prototype. If I can't describe the product convincingly, I'm not ready to build it.
Software engineering5 min read
The API design mistake that haunts every version
I leaked my database shape through my API, and now every internal change is a public breaking change. Don't do this.
Open source5 min read
Open source is a reputation ledger, and it's read by humans
Your contributions form a public record of how you work. Issue comments age better than commits — and employers read both.
Pharmacy × Tech5 min read
What generics and open source have in common
Generic medicines and open source software share a philosophy: knowledge should compound, access should be wide, and trust needs verification.
AI6 min read
AI-generated code has a maintenance problem nobody prices in
Generating code got cheap. Owning it didn't. Every line still has to be understood by someone, eventually, at 2am.
Product building5 min read
The hidden cost of 'we'll add it later'
Technical debt isn't the code you wrote badly — it's the decisions you postponed. Every 'later' collects interest.
Software engineering6 min read
Why I stopped optimizing too early — a story in three builds
Build one: over-engineered and late. Build two: naive and fine. Build three: actually measured, actually fast where it mattered.
Open source5 min read
Why I contribute upstream instead of maintaining forks
Forks feel powerful and rot quietly. Upstream contributions are slower and compound forever. The math I use.
Pharmacy × Tech5 min read
The consultation as a requirements-gathering technique
Pharmacists are trained to extract the real problem from what patients first say. So are good engineers. The technique transfers exactly.
AI6 min read
Testing AI features is different — here's my setup
You can't assert on a model's exact output. You can assert on structure, invariants, and failure behaviour. That's where tests go.
Product building5 min read
Onboarding is a promise you make in the first thirty seconds
Users decide faster than your product loads. What the first screen must do — and the sins I stopped committing.
Software engineering5 min read
Rubber ducks and logs: a solo developer's debugging kit
No teammate to brainstorm with? These are the techniques that substitute: structured logging, bisection, and the duck.
Open source4 min read
The bug report checklist that changed my hit rate
Reports that get fixed share a skeleton. Mine: minimal repro, versions, expectation vs reality, and what I already tried.
Pharmacy × Tech5 min read
Adverse drug reactions and postmortems: the same document
Pharmacovigilance invented the blameless postmortem decades before software. The reporting culture is what software is still building.
Product building5 min read
Naming a product is positioning, not poetry
A good name does a job: it says what shelf the product lives on. My framework for choosing names that work.
Software engineering5 min read
The case for boring technology in student projects
Your project's novelty budget is small. Spend it on the product, not the infrastructure — especially when exams exist.
AI5 min read
Choosing models: a decision tree for people with budgets
The flagship model is rarely the right default. A practical way to pick: capability floor, then cost, then latency.
Open source4 min read
Documenting a project you don't fully understand
The best person to document a concept is the person who just struggled with it. Your confusion is the map.
Pharmacy × Tech5 min read
Compounding formulas and configuration management
A compounding worksheet is a config file with clinical stakes. The discipline of exactness transfers with surprising force.
Product building6 min read
Building products while in school: an honest manual
Exams don't negotiate and neither do servers. The systems I use to build seriously without failing at either thing.
Software engineering6 min read
How I structure a Next.js project so it doesn't rot
Content as data, components as functions of data, and a folder layout that answers questions instead of raising them.
Open source5 min read
The economics of free software work
Nobody pays you to fix bugs at midnight, so the currency must be something else. Skills, reputation, and access — honestly accounted.
AI5 min read
Context windows are a design constraint, not a spec number
Bigger windows changed what's possible — and quietly changed how systems should be designed. Retrieved beats stuffed.
Pharmacy × Tech5 min read
The antidote mindset: designing for reversal
Medicine always asks: what's the antidote if this goes wrong? Software should ask the same question before shipping anything.
Product building4 min read
The changelog is a product's memory and its marketing
A public changelog compounds: users see momentum, you see progress, and 'what changed' stops living in your head.
Software engineering4 min read
What code review taught me about writing prose
Same skills, different medium: clarity, structure, knowing your reader, and killing your favourite sentence.
Open source4 min read
How to ask a maintainer a question they'll want to answer
Good questions show the work already done. The difference between 'how do I do X?' and a question that gets a thoughtful reply.
AI6 min read
What building AI products in Nigeria actually looks like
Payment rails, power, and latency shape every AI product decision here. Constraints don't stop building — they shape what gets built.
Pharmacy × Tech5 min read
Stock-outs and error budgets: inventory meets reliability
A pharmacy that never runs out and an API that never fails are managed the same way: with levels, buffers, and budgets.
Product building5 min read
Working with a distributed team across time zones
Async isn't a compromise — it's the design. What works for collaborating across countries without burning anyone's sleep.
Software engineering4 min read
The one-file rule for keeping side projects alive
If a side project can't be understood from one file after three months away, it's already dead. Design for your returning self.
Open source5 min read
GrantFox, bounties, and getting paid for open source work
Bounty platforms connect contributors with funded issues. What I learned about their reality: scope, proof, and payment rails.
Pharmacy × Tech5 min read
Interviewing patients taught me to interview users
Both patients and users undersell their problems and oversell their compliance. The questioning technique that gets the truth from both.
AI4 min read
Why I keep a prompt journal
A plain file of every prompt that worked, every one that failed, and why. The cheapest AI skill you can build.
Product building5 min read
How I decide what to build next when everything is urgent
Urgency is loud and importance is quiet. The scoring system that keeps my roadmap honest when every idea screams.
Software engineering5 min read
Types are documentation that can't rot
Comments drift, wikis gather dust, but a type error fires the day reality changes. TypeScript changed how I document.
Open source4 min read
Open source as a student: what to contribute with no time
Between lectures and exams there is no spare time — only small time. The contribution formats that fit in an hour.
Pharmacy × Tech5 min read
Why I document like a pharmacist
Clinical notes follow a discipline: write for the next person, timestamp everything, separate observation from interpretation. So should engineering docs.
AI6 min read
AI hallucination: a field guide from someone who ships
Hallucinations aren't random. They cluster in predictable places: numbers, citations, specifics, and anything you'd verify twice.
Product building5 min read
Building with AI without losing your engineering judgement
The convenience trap is real: muscle memory fades when a machine does the lifting. How I keep my own skills sharp.
Software engineering5 min read
Deploying on Friday: a risk framework, not a superstition
'Never deploy on Friday' is a slogan. What you actually need is reversibility, monitoring, and an honest blast radius.
Open source5 min read
The difference between contributing and taking
Opening issues for free labour, demanding features, vanishing after the fix — the patterns that make maintainers close their doors.
Pharmacy × Tech5 min read
Two careers, one person: an identity for the in-between
Neither 'pharmacy student' nor 'developer' holds the whole truth. On building an identity that doesn't require choosing a lane.
Product building6 min read
What a product ecosystem means — and why solo is lonely
One product is a bet. Several products that share users, lessons, and infrastructure are a portfolio of options. The difference is design.
AI5 min read
The best AI feature is the one users don't notice
Generation demos impress. Inference that quietly fills gaps — autocomplete, defaults, cleanup — is what users keep.
Software engineering6 min read
Everything I know about software started in a pharmacy lab
Precision, procedure, verification, and respect for failure modes: the lab is a systems engineering course in a white coat.
Open source4 min read
Why I stopped caring about GitHub stars
Stars measure attention, not usefulness. The repositories I'm proudest of have triple-digit stars and daily users; the ones with thousands taught me nothing.
Pharmacy × Tech5 min read
The clinical eye and the code review
Pharmacists are trained to spot what's subtly wrong on a prescription. The same trained noticing makes code reviews catch what tests miss.
Pharmacy × Tech5 min read
Triage is triage: the emergency shelf and priority queues
Pharmacies triage by consequence, not by noise. Support queues, bug backlogs, and roadmaps should work the same way.
AI7 min read
From notebook to production: the gap nobody blogs about
The demo works. Production is a different country. Version drift, cold starts, monitoring, and the user who types nothing you expected.
Get new posts by email
One email when something worth reading lands. No spam, no schedule.
Every post here is written by me. In a reader instead? Subscribe via the RSS feed. Written by Nzubechukwu Cyprian, Pharmacy student at UNN and full-stack developer.