Veteran dev calls for coding LLMs that natively follow instructions, not just suggest
ivan_bezdomny · x · 2026-09-25
The author argues labs should train coding-focused LLMs from the ground up to natively follow instructions — not treat them as suggestions — with a hierarchy of commands, effectively a programming language for managing software projects via an LLM.
With 20+ years of coding experience, he uses Claude Code and Codex daily and won't go back to hand-coding, but wishes for models that obey specs. His favorite projects remain from the Opus 4.6 era, where he could define requirements and have Opus run unit tests and backups on every update — or at least 90% of them.
More from coding & agent
- SkillRL (NeurIPS 2026): 7B model beats GPT-4o by 41% via recursive skill evolution — cihangxie · 2026-09-25
- Weave Code Max offers $50+ of coding model usage for $10/month — ycombinator · 2026-09-25
- Mad science: Jev autopilot lands a plane in a terminal flight simulator — bilawalsidhu · 2026-09-25
- Greptile reviewed 395K PRs at NVIDIA, cutting merge time from 24h to 6h — ycombinator · 2026-09-25
- Open-source Claude skill turns one prompt into stop-motion claymation films in Blender — angrypenguinPNG · 2026-09-25
- Descript launches MCP to edit video from inside Claude, ChatGPT, and Codex — descript · 2026-09-25