Dirty Runtime
AI wrote polished React code for my dashboard. It compiled. It rendered. And the logic was dead wrong. Debugging AI code is harder because it never hesitates.
Read more →Blog
Pumping for a Fix
Small changes compound. A pretty terminal, a green checkmark, a checked-off task. These tiny dopamine hits aren't distractions. They are the fuel.
Read more →Nothing to Penetrate
An AI lab called its new model too dangerous for public release. It scanned a codebase hardened by decades of discipline and found nothing to penetrate.
Read more →Bare Promises
AI agents need purpose-built tooling. But calling a language agent-first without studying how LLMs interact with code turns a hypothesis into marketing.
Read more →Hooked by Design
AI coding tools run on the same psychological mechanism as slot machines. A friend of mine learned this the hard way: five agents, no sleep, complete burnout.
Read more →Three F's to Get You Off the Market
Most engineers grind LeetCode and polish resumes. The truly elite ones found a faster path. It starts with doing something because it feels good.
Read more →Post-AI Clarity
AI wrote a feature set in two weeks. A lead engineer added integration testing. The result shipped faster than any human team, and it was bulletproof.
Read more →Overdressed and Underperforming
I told everyone to pack their AGENTS.md with every detail. A recent study proved me wrong. More instructions make AI agents perform worse. Here is why.
Read more →Knock Before You Commit
AI coding agents will silently guess your missing requirements and ship the wrong thing. The fix is not better prompts. It is making them ask before they build.
Read more →Your Fundamentals Are Showing
AI generates code faster than anyone can read it. The result is comprehension debt. Smart teams solve it by pairing AI speed with proven fundamentals.
Read more →Disclaimer: All content reflects my personal views only and does not represent the positions, strategies, or opinions of any entity I am or have been associated with.