Karpathy shares how to make AI outputs like aircraft manuals
Karpathy, a former OpenAI member now at Anthropic, posted on X about using language models to generate text in ASD-STE100, a controlled English standard for aircraft maintenance manuals. He found the format—with its strict rules and limited vocabulary—produces clearer, more structured output than unconstrained writing. The standard, maintained by European aerospace firms, restricts writing to…
Key points
- Karpathy tested ASD-STE100, a strict English standard for technical manuals, to improve AI output clarity
- Open-source tool applies STE rules but lacks the full dictionary, per its GitHub README
- Karpathy recommends diagrams, HTML, and explainer videos as clearer AI output formats
Karpathy also suggested alternative formats for model outputs: diagrams, HTML pages, and custom explainer videos. He highlighted the potential of videos, referencing the style of 3Blue1Brown math tutorials, and noted that narration could be generated via ElevenLabs or free local alternatives. His post emphasized shifting focus from manual output generation to oversight and higher-level tasks as AI handles more routine work.
OpenAI Founding Member: Make LLMs Write Like An Aircraft Manual via @sejournal, @MattGSouthern
Search Engine Journal · 2 October 2026
Loading the full article…
This text was published by Search Engine Journal and written by Matt G. Southern. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researcher tests GPT-6 Astra in WoW with agent-wow framework · 1 src
- GitHub releases KernelOPT for agentic GPU kernel optimization · 1 src
- Allen Institute for AI open-sources AstaBrief 8B for faster scientific reports · 1 src
- Datalab releases OmniExtractBench to audit PDF extraction accuracy · 1 src
- What are world models and how do they predict environments? · 1 src
Comments
via GitHub Discussions