Greg

@gmotyl.bsky.social

Software Engineer - focused on frontend and AI

Should we even read all this generated code? Blunt test: if you had time to read everything you generated last week, you're not generating enough. Engineering shifted — from handcrafting line by line to designing the tools, harnesses and loops. Cheap, for now. Someone pays that bill.

motyl.dev

New issue of my newsletter is out. This week I got stuck on one question: if AI writes the code, what's left for me? Turns out a lot of it is just reviewing until I stop actually reading. A few reads that pushed my thinking on this:

motyl.dev

New newsletter issue is up. Short and sharp this week. The agent tooling keeps sprinting — cloud agents, subagent primitives, GLM-5.2 vs Kimi head-to-heads — but the best reads pulled the other way: stay skeptical, don't over-engineer. Build fast, keep your judgment on. 👇

motyl.dev

The mood among engineers is shifting: we're done pretending you can vibe-code and walk away. This week — Claude Code vs Antigravity vs Codex evaluated, "Loop Engineering," and six months of data on how AI search killed keyword SEO.

motyl.dev

Issue #10 of motyl.dev Weekly is out. The theme: agents are software again. Eval-ops, harnesses, compiler-style output checks, mission-control dashboards. Plus my own deep dive on Pavilio, the open-source panel I built to keep multiple CLI agents in line. https://motyl.dev/newsletter/10

I was about to vibe-code my own Postman alternative. Then KIVO showed up — open source, lightweight, free. Plus how to structure CLAUDE.md after 1000 sessions, Railway moving off Next.js, and the operational stuff people are finally writing down about agent dev.

motyl.dev

Two major npm security incidents in one week — Axios got a trojan, Claude Code source leaked via source maps. Plus: TypeScript 7.0's Go rewrite, Kent Beck on TCR for AI agents, and GitHub's AI-era dominance questioned. New issue of my weekly newsletter: motyl.dev/newsletter/5

motyl.dev