Writing about what I build

Vibe coding treats AI as an oracle you gamble on. AI-Orchestrated Engineering treats it as an instrument you conduct. Here's the discipline I've been building and why the distinction matters.

A living reference for writing posts thumbnails, code, tables, alerts, footnotes, images, audio, video, and YouTube embeds.

How Ken Perlin's gradient noise creates infinite terrain, clouds, and fire and why Simplex improved it thirty years later.

A frontend engineer's guide to the architecture behind GPT self-attention, positional encoding, and the encoder-decoder split.

Render infinite geometry on a fullscreen triangle SDFs, sphere tracing, and the shader loop that replaced polygon soup.

How to ship a markdown blog with zero database, full SEO, and static export the same stack powering mahdi.is-a.dev.

Loss surfaces, learning rates, momentum, and Adam the optimization loop that makes neural networks actually learn.

Lessons from building NPUI tokens, primitives, and the gap between Figma and production.

The horse is doing the running. You're still the one who has to stay balanced, read the terrain, and not fall off. That's the whole difference between vibe coding and actually directing the thing.

I taught our building's maintenance guy how to vibe code for fun. Twenty minutes later he'd chatted his way into a working feature inside one of my real apps. Here's what actually happened, and why it didn't disprove anything I've written in this series.

A five-level, self-scoring checklist to find out honestly whether your AI-assisted workflow is AI-Orchestrated Engineering or vibe coding with extra steps.

Pipeline, fan-out/fan-in, adversarial verification, judge panels the distributed-systems patterns that make AI-Orchestrated Engineering practical instead of aspirational.

Intent, Decomposition, Delegation, Verification, Synthesis the framework I use to turn 'AI writes code fast' into 'AI-assisted systems are correct.'

Vibe coding feels fast because it defers all its cost to later. A field guide to the five failure modes I see over and over and why they're structural, not incidental.

How React 18 schedules updates, keeps the UI responsive during heavy renders, and what useTransition actually buys you.

The data structure behind PostgreSQL, SQLite, and MySQL indexes balanced trees, page alignment, and why O(log n) isn't enough.

From Figma's design engine to Cloudflare Workers why WASM is becoming the portable bytecode layer of the web stack.

How distributed systems agree on a log when nodes fail terms, heartbeats, and why Paxos stayed in papers while Raft got textbooks.

Why implicit flow died, how PKCE protects public clients, and the token flow every React app should implement.

Git isn't a filesystem with history it's a content-addressed object database with labels. Once you see the DAG, everything clicks.

Building Analytick taught me that you can have real-time dashboards and respect user privacy if you design constraints in from day one.