
Bring Back Manual QA: AI Fixed What Made It Painful
Your team ships AI-generated code faster than anyone can check it. Manual QA should come back, as a senior role that owns what ships.
Articles on software development, infrastructure, and building effective teams. No fluff, just practical experience from shipping production code.
11 articles

Your team ships AI-generated code faster than anyone can check it. Manual QA should come back, as a senior role that owns what ships.

A plan I'd reviewed and approved still hid two design problems. These two Claude Code skills found them and took the clutter out of my head.

Re-prompting an LLM for the same deterministic task means paying again for work you already solved. Use the model once to write a script, then run it for free.

Your team is busy and nothing ships. The cause usually sits in the translation layer between business need and technical execution, and almost nobody owns it.

The OpenClaw CLI boots Node.js on every terminal open just to print static completions. A cache that refreshes itself removes the 3.7-second hit.

Five minutes of profiling found 5 seconds of wasted shell startup. Caching, lazy-loading and deferred init brought it down to about 1 second.

How I stopped juggling port numbers in my monorepo with Caddy wildcard routing. Every app gets a memorable name like pregni-web.dev.localhost.

Cursor made AI coding mainstream, then broke developer trust. Why it happened, and what works better for teams shipping production code.

Why throwing prompts at ChatGPT doesn't make your team better, and how applying engineering discipline to AI-assisted development actually does.

How we built a 7-step client discovery form with zero JavaScript frameworks, session-based resumability, and type-safe backend error handling in 8 hours.

Import your UI library everywhere and a redesign, or a move from MUI to Mantine, touches hundreds of files. A thin facade keeps it a small change.