Ask HN: 寻找适用于编程的 LLM 概念模型资源？
1 分•作者: pramodbiligiri•5 个月前
我正在努力从概念上充分理解 LLM，以便能够预测它们在生成代码方面的能力（和局限性）。这是一个合理的目标吗？有什么好的资源吗？
到目前为止，我查阅了：
1. 《Vibe Coding》，作者：Steve Yegge 和 Gene Kim (https://www.amazon.in/Vibe-Coding-Building-Production-grade-Software/dp/1966280025)。这本书有一些实际的例子和许多指导方针。但它没有太多理论，而且据我所知，它也没有从概念上解释 LLM。
2. 《从零开始构建 LLM》，作者：Sebastian Raschka (https://www.manning.com/books/build-a-large-language-model-from-scratch)。这本书看起来很深入。但我实际上并不想 <i>构建</i> 一个 LLM。
3. 《AI Engineering》，作者：Chip Huyen (https://www.amazon.in/AI-Engineering-Building-Applications-Foundation/dp/1098166302)。这本书看起来很有希望，尽管它不是以代码为中心的。
也许类似《Claude 代码的工作原理》（https://code.claude.com/docs/en/how-claude-code-works）这样的内容，但需要更详细地阐述。
谢谢。
查看原文
I am trying to understand LLMs conceptually well enough to be able to predict their capabilities (and limitations) when it comes to generating code. Is that even a sensible goal? Are there good resources?<p>So far I've looked at:<p>1. Vibe Coding, by Steve Yegge and Gene Kim (https://www.amazon.in/Vibe-Coding-Building-Production-grade-Software/dp/1966280025). This has some practical examples and many guidelines. But there is not much theory and this does not explain LLMs conceptually AFAICT.<p>2. Build an LLM from Scratch, by Sebastian Raschka (https://www.manning.com/books/build-a-large-language-model-from-scratch). Seems in-depth. But I don't really want to <i>build</i> an LLM.<p>3. AI Engineering, by Chip Huyen (https://www.amazon.in/AI-Engineering-Building-Applications-Foundation/dp/1098166302). This seems promising, although it is not coding focussed.<p>Perhaps something like How Claude Code Works (https://code.claude.com/docs/en/how-claude-code-works) but fleshed out in more detail.<p>Thanks.