Karpathy says Codex and Claude Code need user control over subagent roles and model choice
emollick · x · 2026-07-22
Andrej Karpathy says Codex and Claude Code need finer control over subagents
Andrej Karpathy points out a practical limitation in tools like Codex and Claude Code: users need more control over how orchestrator AIs configure their subagents.
His core point is that the human should be able to decide:
- whether a subagent should do research, writing, or user testing
- which model each subagent should use
Without that control, the system becomes “a router problem again” — meaning the orchestrator is still making too many hidden delegation decisions on the user’s behalf.
More from coding & agent
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- Treating agents like 50 First Dates: a 3-layer context system so every conversation doesn't start from zero — evielync · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- ARRM targets silent economic regressions in AI agents that functional tests miss — Beautiful_Belt_601 · 2026-09-11
- Dev builds browser 3D pizza delivery game with Claude: physics, GPS pathfinding, traffic AI — vinishkapoor · 2026-09-11
- Build X Carousel Posts from One Wide Image: A Splitter Tool Plus YouMind Skill Workflow — sujingshen · 2026-09-11