4 min read DPS

Daily Productive Sharing 1259 -AI Skeptics Are All Nuts

Daily Productive Sharing 1259 -AI Skeptics Are All Nuts
Photo by Marek Piwnicki / Unsplash

One helpful tip per day:)

Thomas Ptacek believes that large language models (LLMs) understand code better than we do:

  1. Agents can autonomously navigate your codebase, write files, run tools, compile code, execute tests, and iterate based on the results.
  2. LLMs can write most of the repetitive code you’d otherwise need to write—and most code in real-world projects is repetitive.
  3. You are still responsible for whatever gets merged into main. That was true five years ago, and it’s still true tomorrow—whether or not you use LLMs.
  4. If you're building something others will rely on using LLMs, read the code. In fact, you’ll probably need to do even more than that.
  5. He enjoys muttering to himself while moving code around, deleting dumb comments and chuckling—because code still needs to be read line by line.
  6. That’s why he likes Zed’s agentic mode: it lets the agent do its work in another tab, then notifies you when it's done.
  7. Everyone writes mediocre code sometimes. And often, mediocre code is good enough—not all code is mission-critical.
  8. Even if LLMs only ever produce mediocre code, that’s still a massive win—because it means humans don’t have to write that mediocre code themselves.
  9. LLMs _will_ likely replace many software developers. This isn’t some abstract issue—it mirrors the economic disruptions faced by other industries over the past 30 years.

If you enjoy today's sharing, why not subscribe

Need a superb CV, please try our CV Consultation

Thomas Ptacek 认为大语言模型对代码的理解比我们更深入:

  1. 智能体可以自主浏览你的代码库，直接编写文件，运行工具，编译代码，执行测试并根据结果进行迭代。
  2. 大语言模型可以编写你需要写的大部分重复性代码。而大多数项目中的代码，恰恰是重复性的。
  3. 你对合入 main 分支的内容一直都负有责任。五年前如此，明天也一样，不管你是否使用大语言模型。
  4. 如果你用大语言模型构建了别人会依赖的东西，那就读一下代码。事实上，你可能还得做得更多。
  5. 他喜欢边移动代码边嘀咕，然后一边删掉那些愚蠢的注释一边偷笑。反正代码还是得一行行看。
  6. 这就是他喜欢 Zed 的智能体模式 的原因：它会让你点开别的标签页，让它自己干活，干完后会发桌面通知提醒你。
  7. 每个人都写过平庸的代码。平庸的代码：很多时候也挺好。不是所有代码都那么重要。
  8. 即使大语言模型最终只会写出平庸代码，这本身也是巨大的进步——人类就少写了那么多平庸代码。
  9. 大语言模型确实可能会取代很多软件开发者。这不是我们可以高高在上的问题。我们的工作岗位和其他行业过去三十年所面临的冲击并无不同。

如果你喜欢的话，不妨直接订阅这份电子报 ⬇️

Dr Selfie

Dr Selfie