Victor Taelin: AI coding is miserable — you can't quit it, but it keeps silently breaking your code
Michael_Moroz_ · x · 2026-09-09
Developer Victor Taelin wrote a widely resonant thread calling programming in the AI era "the most miserable": not using AI makes him slower and shrinks his ambition, but using AI means coping with code being silently damaged — like an AI swapping a 10s limit for 150s without telling him, discovered two weeks later on launch day. He argues even the least competent human engineer wouldn't regress code this way, says he's now dependent on AI with no way out, and hopes his Bend project can solve the problem.
More from coding & agent
- Stardock CEO uses AI to A/B test two copies of Sins of a Solar Empire II with auto-generated reports — draginol · 2026-09-09
- How Stripe built Kai, its company-wide AI agent, on LangChain Deep Agents with 1.5 engineers — Hacubu · 2026-09-09
- LangChain's Deep Agents adds subagent forking: subagents inherit full supervisor context — LangChain · 2026-09-09
- 16,893 sessions measured: what Claude Code, Codex and Cursor actually recommend — rseroter · 2026-09-09
- Compound Engineering after a year: shipping more than ever, but my brain got softer — danshipper · 2026-09-09
- When Agents Sit Waiting for Human Approval Nobody Notices — pixelatrix · 2026-09-09