Karpathy：让 LLM 用受控语言写作、画图表、做网页、生成讲解视频 (2026)

中文译文

我们以后会花多得多的时间，去理解语言模型的输出。这里是一些想法、技巧和窍门：1

写作。有一个我用下来效果不错的做法：让 LLM 用 ASD-STE100 来解释一件事。ASD-STE100 是一套受控语言规范，最初是为了编写航空航天维修文档而开发的。LLM 对这套语言很熟悉，它自带一整套严格的约束，对干净的写作文风限制很多，我常常觉得这样写出来的东西反而好读得多。有时候我会试着把它放松一点，比如要求「做到 ASD-STE100 的八成」，因为这套规范确实相当严苛。不过还有更好的办法：

图表／图像。与其写文字，不如让 LLM 画一张示意图。图往往更容易处理、解析和理解。不过还有更好的：

网页。要求「用 HTML 输出」，就能得到一张漂亮、可交互的网页。LLM 的前端能力越来越好，可以做出好看的体验、动画等等。不过还有更好的：

讲解视频。我最看好的输出格式，是针对任意主题生成的完全定制／量身定做的讲解视频。可以拿这类指令做实验：「用 3b1b 的风格，做一段关于 X 的讲解视频。用我的 ElevenLabs API key 来做音频旁白。」（后者需要你有 API key，或者你也可以让 LLM 帮你找一些不错的免费替代方案，用你自己的本地算力跑。）这件事真的开始能用了！

总结：

随着 LLM 变得越来越强，它们会越来越多地自己完成那些基础工作，我们工作中很大一部分会上移到抽象层，变成监督和理解。
幸运的是，LLM 在这件事上也能帮上忙：随着智能和代码越来越充裕，你可以要求生成又大又定制、用完即弃的软件产物（比如网页应用、讲解视频），而放在以前，做这些东西根本不值得。在这个方向上把边界往外推一推，你会被结果惊到。

英文原文

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. 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:
Diagrams / 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:
Web 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:
Explainer 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!
In summary:
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.
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.

原帖发布于北京时间 2026 年 10 月 2 日 08:37:00。1

Karpathy 的原帖：

Loading content card…

References

1
Karpathy 原帖
x.com

中文译文

我们以后会花多得多的时间，去理解语言模型的输出。这里是一些想法、技巧和窍门：1

图表／图像。与其写文字，不如让 LLM 画一张示意图。图往往更容易处理、解析和理解。不过还有更好的：

网页。要求「用 HTML 输出」，就能得到一张漂亮、可交互的网页。LLM 的前端能力越来越好，可以做出好看的体验、动画等等。不过还有更好的：

总结：

随着 LLM 变得越来越强，它们会越来越多地自己完成那些基础工作，我们工作中很大一部分会上移到抽象层，变成监督和理解。

幸运的是，LLM 在这件事上也能帮上忙：随着智能和代码越来越充裕，你可以要求生成又大又定制、用完即弃的软件产物（比如网页应用、讲解视频），而放在以前，做这些东西根本不值得。在这个方向上把边界往外推一推，你会被结果惊到。

英文原文

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:

Writing. 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:

Diagrams / 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:

Web 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:

Explainer 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!

In summary:

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.
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.

原帖发布于北京时间 2026 年 10 月 2 日 08:37:00。1

Karpathy 的原帖：

Loading content card…

Karpathy：让 LLM 用受控语言写作、画图表、做网页、生成讲解视频

中文译文

英文原文

References

Related content

Karpathy：让 LLM 用受控语言写作、画图表、做网页、生成讲解视频

中文译文

英文原文

References

Related content