Andrej Karpathy: rise up the output-format ladder (ASD-STE100 to bespoke explainer videos)

ID: e3887eee
原文链接: https://x.com/karpathy/status/2105819303471976479
作者: Andrej Karpathy
日期: 2026-10-02
分类: learning
来源类型: x_post
标签: llm-output、asd-ste100、explainer-video、prompting
质量评分: 3/5
抓取时间: 2026-10-02 (daily-intake-evening, opencli/web fetch)

中文导读

Karpathy 10 月 2 日的推文（发布当天即 11k 赞）给出一道递进的 LLM 输出格式阶梯：先让 LLM 用 ASD-STE100（航空维护文档的受控英语规范）写解释，嫌太硬就要求「80% 程度的 ASD-STE100」；再往上是让模型画图；再往上是直接生成 HTML 交互网页（前端能力已足够做动画）；他最看好的顶端是「为具体问题定制的解释视频」——例如要求 3b1b 风格动画配 ElevenLabs 配音，整套产物是一次性、定制化、过去根本不值得做的「软件废料」。他的总结：LLM 越强就越会自己干完脏活，工程师的工作沿抽象链上移到理解与监督，而 LLM 也在放大这一步，因为智能与代码充裕后，一次性网页与定制视频反而成了最优输出格式。

为什么值得关注

输出格式阶梯：受控英语到图到 HTML 到定制解释视频，每升一档可读性上一个台阶；顶端产物是过去不值得做的一次性软件。

English Summary

Karpathy's Oct 2 tweet (11k likes within hours) lays out a progressive ladder of LLM output formats: have the LLM explain in ASD-STE100, the controlled-English spec from aerospace maintenance docs (soften to '80% of the way to ASD-STE100' when too stringent); then diagrams instead of prose; then full HTML web pages with interactivity and animation; and at the top, what he is most bullish on - fully bespoke explainer videos for any arbitrary topic, e.g. a 3b1b-style animation with ElevenLabs narration: one-off, customized 'software artifacts' that were never worth producing before. As LLMs do more of the legwork autonomously, human work rises up the abstraction ladder into oversight - and abundant intelligence makes disposable pages and custom videos the optimal output format.

Obsidian 原文摘录（抓取正文头部）

[
  {
    "id": "2105819303471976479",
    "author": "karpathy",
    "bio": "",
    "text": "We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:\n\nWriting. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for \"80% of the way to ASD-STE100\" because the spec is quite stringent. But even better:\n\nDiagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:\n\nWeb pages. Ask for output \"in HTML\" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:\n\nExplainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like \"Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration\". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!\n\nIn summary:\n- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.\n- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.",
    "likes": 30930,
    "retweets": 3362,
    "created_at": "Fri Oct 02 00:37:00 +0000 2026",
    "url": "https://x.com/karpathy/status/2105819303471976479",
    "has_media": true,
    "media_urls": [
      "https://pbs.twimg.com/media/HTlaHqgbwAAS1lv.png"
    ],
    "card": null,
    "quoted_tweet": null
  },
  {
    "id": "2106046406360543668",
    "author": "datfreedom",
    "bio": "My opinions are my own",
    "text": "@karpathy This is what I have been doing la

Notes

Content grounded in a same-day fetch (opencli web read / twitter thread / direct HTTP) during daily-intake-evening.
中文导读 block is the entry's summary_zh (verbatim); 为什么值得关注 is the entry's one_liner; no claims beyond the fetched source text were added.
Source file cached at /tmp/aaif-evening/karpathy.json during the run.