BUILD / SHIP / LEARN

The Lab

Real products. Real process. This is what it looks like when I actually build something.

What I've built

Products I've taken from idea to production — with real users, real infrastructure, real consequences.

Live

AudioByte

AI-powered audio learning for university students. Converts course materials into voice-native study sessions — built for students in African universities who learn better by listening.

TypeScript React Supabase Python TTS AI
Live

Memoir

This site. A personal space to think, document ideas, and build in public. Built as a showcase of process and output — not just a blog.

Astro TypeScript React
Next up

Something new

Always something in the pipeline. Follow along on The Venture for updates as things take shape.

How I actually build

Not vibing, not prompting randomly. A structured engineering process built on real fundamentals — packaged as named, reusable agent skills.

01
/grill-with-docs

Align first

Before writing a line of code, I interview myself and the problem relentlessly. Every branch of the decision tree gets resolved. The codebase gets a shared language — no jargon drift, no miscommunication.

02
/to-prd → /to-issues

Plan with precision

The aligned understanding becomes a PRD. The PRD breaks into vertical slices — tracer bullets through every layer (schema, API, UI, tests). Each slice is independently demoable before the next starts.

03
/tdd

Red → green → refactor

Test-driven development with a strict feedback loop. A failing test is written first. Then the fix. Then the refactor. This isn't dogma — it's the only way to give an AI agent consistent feedback that produces reliable code.

04
/diagnose

Debug like a scientist

Reproduce → minimise → hypothesise → instrument → fix → regression test. No random changes, no "just try this." Every bug gets a loop that ends with a test so it can't come back.

05
/improve-codebase-architecture

Fight entropy deliberately

AI accelerates software entropy. Running this skill regularly finds deepening opportunities — places where shallow modules can become deep ones. Design investment every day, not just at the start.

These are real skills — not a framework I invented. They live in .skills/ in this repo, sourced from mattpocock/skills and adapted to how I work.

What's in the box

Languages

TypeScript Python SQL

Frontend

Astro React TailwindCSS

Backend / Infra

Supabase Edge Functions REST APIs

Agents & AI

Cursor Claude Antigravity TTS / STT APIs

Workflow

Notion GitHub skills.sh

Want to see what I'm doing right now?

Open /now