返回列表
🧠 阿头学 · 🪞 Uota学 · 💬 讨论题

AI 正在把开源理想变成“可塑的计算机”

这篇文章的判断是对的半截:AI 确实显著降低了改造软件和系统的门槛,但 DHH 把早期极客现象过度包装成普遍趋势,而且明显在为 Linux/Omarchy 叙事加码。
打开原文 ↗

2026-04-16 原文链接 ↗
阅读简报
双语对照
完整翻译
原文
讨论归档

核心观点

  • AI 把“可修改”从权利推向能力 作者抓住了一个真问题:开源长期提供的是法律和技术上的修改权,但大多数人实际上用不上;AI 现在确实在压缩读代码、找修改点、写补丁、调配置的成本,这一步是实质进展,不是口号。
  • Linux 会比封闭平台更先吃到红利 这个判断基本成立,因为系统级可塑性最终取决于权限、源码和接口开放度;在这点上 Linux 的优势是真实的,Windows/macOS 虽然也能脚本化和外挂式定制,但深度改造能力明显弱一档。
  • 作者把“能改一次”说成了“能长期掌控” 这是全文最大的偷换。AI 帮用户改出功能,不等于用户就获得了稳定、安全、可维护的系统主权;真正难的往往是升级兼容、故障排查、回滚治理,而不是第一次生成代码。
  • Omarchy 案例有启发,但证据远远不够 作者用自己社区里的成功样本来论证未来方向,这可以说明趋势萌芽,但不能证明大众市场会快速跟进;极客用户的容错率和动手意愿,本来就和普通用户不是一个层级。
  • 这是一篇愿景文章,不是一篇严谨分析 DHH 的表达有煽动力,也有真实洞察,但结论明显超出了证据范围;尤其“黑箱计算机会迅速过时”的判断,现在看更像立场宣示,而不是已被验证的产业事实。

跟我们的关联

1. 对 ATou 意味着什么:未来不该只把 AI 当内容助手或流程助手,而要把“让产品被用户和 AI 二次塑形”当成一条产品战略;下一步可以直接审视现有产品有哪些层是可配置、可脚本、可暴露接口的,哪些层仍是黑箱。 2. 对 Neta 意味着什么:判断一个 AI 产品有没有长期价值,不能只看模型能力,要看它是否站在开放底座上;下一步可以用“权限开放度—修改成本—维护闭环”这个框架筛项目,而不是只看 demo 是否惊艳。 3. 对 Uota 意味着什么:这篇文章强化了“AI 时代的个人系统不该只是被动使用,而该能按自己心智重构”的方向感;下一步可以从最小闭环开始,不是先重做操作系统,而是先用 AI 改一个本地工具、一个工作流、一个界面层。 4. 对通用判断意味着什么:AI 的真正杠杆可能不是替你操作软件,而是替你改造软件;下一步讨论时应把焦点从“agent 会不会点按钮”转到“agent 能不能稳定改系统并负责维护”。

讨论引子

1. AI 降低的是“首次改造门槛”,还是已经足以降低“长期维护门槛”? 2. Linux 会因为 AI 重新获得主流个人计算入口,还是依然只会停留在高意愿用户圈层? 3. 产品应该追求一致性和封闭控制,还是应该主动让用户和 AI 改自己?

可塑的计算机

开源曾经许诺,用户可以自由修改自己正在运行的任何代码。然而现实是,几乎没有人真的这么做过,因为这实在太难了。如今,有了 AI,这件事突然不再难了。

这非常令人兴奋。能够给任何本地的开源应用添加功能,然后使用这个为自己量身定制的分支,这朝着开源最初的承诺迈出了不可思议的一步。

这也不只是普通用户的事。即便你是程序员,也未必熟悉那个应用所使用的编程语言。即便你熟悉,花时间理解任何一个有一定规模的代码库,也都不是一件轻松的事。AI 正在压缩这种复杂性,并以惊人的速度让它变得可塑。

不过,更让我兴奋的是,当这种能力被用到操作系统上,进而用到整台计算机上。那时你能改变的不只是单个应用,而是系统菜单栏、窗口管理器、通知系统,以及几乎一切东西。

但你只能在 Linux 上做到这一点。在 Windows 和 macOS 上,操作系统的核心元素属于制造它们的公司。虽然通常也可以对某些方面做一些 hack,但距离 Linux 所允许的那种真正可塑的计算机,还差得很远。

我已经在 Omarchy 的世界里多次看到这种情况:那些并不特别懂技术的用户,在 AI 的帮助下把系统改造成自己的样子,并对结果感到由衷欣喜。

虽然这件事现在仍然相当极客,但我不认为它会长期局限在这个小圈子里。随着模型变得更强大,认为你的系统只能是一个固定黑箱的想法,很可能会迅速变成一种过时观念。

一如既往,未来已经到来,只是分布得并不均匀。

关于 David Heinemeier Hansson

作为 37signals 的共同所有者兼 CTO,为弱势一方打造了 BasecampHEY。创建了 Ruby on RailsHotwireKamalOmarchy。写过 REWORKIt Doesn't Have to Be Crazy at WorkREMOTE。作为一名赛车手赢得过勒芒赛事。投资了丹麦初创公司

订阅即可通过电子邮件收到未来文章,或者获取 RSS feed

The malleable computer

可塑的计算机

Open source promised that users would be free to change whatever code they were running. The reality, however, is that hardly any of them ever did — it was simply too hard. Now, with AI, it suddenly isn't.

开源曾经许诺,用户可以自由修改自己正在运行的任何代码。然而现实是,几乎没有人真的这么做过,因为这实在太难了。如今,有了 AI,这件事突然不再难了。

This is very exciting. Being able to add features to any local open-source application and then use that bespoke fork for your own benefit is an incredible step toward the original open source promise.

这非常令人兴奋。能够给任何本地的开源应用添加功能,然后使用这个为自己量身定制的分支,这朝着开源最初的承诺迈出了不可思议的一步。

This isn't just about regular users, either. Even if you are a programmer, you might not be familiar with the language the application is written in. And even if you are, taking the time to get familiar with any substantial codebase is a tall order. AI is compressing that complexity and making it malleable at a ferocious rate.

这也不只是普通用户的事。即便你是程序员,也未必熟悉那个应用所使用的编程语言。即便你熟悉,花时间理解任何一个有一定规模的代码库,也都不是一件轻松的事。AI 正在压缩这种复杂性,并以惊人的速度让它变得可塑。

What excites me even more, though, is when this power is applied to the operating system, and thus the entire computer. When you're able to change not just individual applications, but your system's menu bars, your window manager, your notification system, your everything.

不过,更让我兴奋的是,当这种能力被用到操作系统上,进而用到整台计算机上。那时你能改变的不只是单个应用,而是系统菜单栏、窗口管理器、通知系统,以及几乎一切东西。

But you can only do this on Linux. With Windows and macOS, the core elements of the operating system are owned by the companies that make them. While it's often possible to hack certain aspects, it's far from truly having the malleable computer that Linux allows.

但你只能在 Linux 上做到这一点。在 Windows 和 macOS 上,操作系统的核心元素属于制造它们的公司。虽然通常也可以对某些方面做一些 hack,但距离 Linux 所允许的那种真正可塑的计算机,还差得很远。

I've already seen this a lot in the Omarchy world: users who aren't super technical making the system their own with the help of AI and being utterly delighted by the outcome.

我已经在 Omarchy 的世界里多次看到这种情况:那些并不特别懂技术的用户,在 AI 的帮助下把系统改造成自己的样子,并对结果感到由衷欣喜。

And while this is still a pretty nerdy thing to do, I don't think it will remain contained to that niche for long. As models get even more powerful, the idea that your system is tied down as a fixed black box is likely to become an archaic notion pretty quickly.

虽然这件事现在仍然相当极客,但我不认为它会长期局限在这个小圈子里。随着模型变得更强大,认为你的系统只能是一个固定黑箱的想法,很可能会迅速变成一种过时观念。

As always, the future is already here, it's just not evenly distributed.

一如既往,未来已经到来,只是分布得并不均匀。

About David Heinemeier Hansson

关于 David Heinemeier Hansson

Made Basecamp and HEY for the underdogs as co-owner and CTO of 37signals. Created Ruby on Rails, Hotwire, Kamal, Omarchy. Wrote REWORK, It Doesn't Have to Be Crazy at Work, and REMOTE. Won at Le Mans as a racing driver. Invested in Danish startups.

作为 37signals 的共同所有者兼 CTO,为弱势一方打造了 BasecampHEY。创建了 Ruby on RailsHotwireKamalOmarchy。写过 REWORKIt Doesn't Have to Be Crazy at WorkREMOTE。作为一名赛车手赢得过勒芒赛事。投资了丹麦初创公司

Subscribe to get future posts via email (or grab the RSS feed)

订阅即可通过电子邮件收到未来文章,或者获取 RSS feed

The malleable computer

Open source promised that users would be free to change whatever code they were running. The reality, however, is that hardly any of them ever did — it was simply too hard. Now, with AI, it suddenly isn't.

This is very exciting. Being able to add features to any local open-source application and then use that bespoke fork for your own benefit is an incredible step toward the original open source promise.

This isn't just about regular users, either. Even if you are a programmer, you might not be familiar with the language the application is written in. And even if you are, taking the time to get familiar with any substantial codebase is a tall order. AI is compressing that complexity and making it malleable at a ferocious rate.

What excites me even more, though, is when this power is applied to the operating system, and thus the entire computer. When you're able to change not just individual applications, but your system's menu bars, your window manager, your notification system, your everything.

But you can only do this on Linux. With Windows and macOS, the core elements of the operating system are owned by the companies that make them. While it's often possible to hack certain aspects, it's far from truly having the malleable computer that Linux allows.

I've already seen this a lot in the Omarchy world: users who aren't super technical making the system their own with the help of AI and being utterly delighted by the outcome.

And while this is still a pretty nerdy thing to do, I don't think it will remain contained to that niche for long. As models get even more powerful, the idea that your system is tied down as a fixed black box is likely to become an archaic notion pretty quickly.

As always, the future is already here, it's just not evenly distributed.

About David Heinemeier Hansson

Made Basecamp and HEY for the underdogs as co-owner and CTO of 37signals. Created Ruby on Rails, Hotwire, Kamal, Omarchy. Wrote REWORK, It Doesn't Have to Be Crazy at Work, and REMOTE. Won at Le Mans as a racing driver. Invested in Danish startups.

Subscribe to get future posts via email (or grab the RSS feed)

📋 讨论归档

讨论进行中…