Karpathy: From controlled-language docs to bespoke explainer videos, better ways to read LLM output
karpathy · x · 2026-10-02
Karpathy shares an escalating set of tips for understanding LLM outputs: ask for ASD-STE100 (an aerospace controlled-language spec) for cleaner writing; request diagrams; ask for full interactive HTML pages; and — his most bullish pick — fully bespoke explainer videos (e.g. "3b1b style video on X, with ElevenLabs narration"). His thesis: as models improve, humans move up to oversight, and cheap intelligence enables custom, discardable software artifacts that never made sense to build before.
More from AGI Musings
- User research: people don't want personal agents to automate everything — the process is the point — sebpaquet · 2026-10-02
- Robin Hanson: LLMs' default worldview comes from an Internet that punished heterodoxy — RichardMCNgo · 2026-10-02
- Researcher: arXiv flooded with slop, ML academia increasingly marginalized — joshua_saxe · 2026-10-02
- "Bad art is worse than any tyrant, but bad AI art is merely tedious"—AI art debate — SydSteyerhart · 2026-10-02
- Robotics papers on arXiv hit 210 per day, up from 20-30 four years ago — DJiafei · 2026-10-02
- DeepMind researcher: billions of recursive agents will soon run over mathematicians — CsabaSzepesvari · 2026-10-02