Asked ChatGPT to polish notes, it went ahead and did the coding work itself
dfinke · x · 2026-09-20
Developer dfinke asked ChatGPT to polish rough notes about adding 3 AI providers to his PSAISuite project using GitHub issues + Copilot — instead, the model went ahead and did the work itself.
A notable agent autonomy example: rather than just editing copy, the model executed the actual engineering task, showing how a plain development plan can be handed to an AI agent and shipped.
More from coding & agent
- Using fast decision models for evals: 200ms judge beats slow LLM-as-judge — holdenmatt · 2026-09-20
- Jev-align: a ~$0.003 alignment gate that scores LLM replies and agent plans before you run them — johnseach · 2026-09-20
- Vite core member finds three methodology flaws in OJ's memory benchmarks vs Vite — cnakazawa · 2026-09-20
- Sanctorum runs AI agent teams in a pixel office with seat-derived permissions — AlexanderJAnderson · 2026-09-20
- How to build an AI second brain in an afternoon: a six-step agent workflow — VibeMarketer_ · 2026-09-20
- SnapBridge plugin one-clicks Chrome screenshots into Google's Antigravity IDE — AI_Andrew · 2026-09-20