TLDRocket
Sign in

Devtools must be open source (exe.dev)

Simon Willison Simon Willison Covered by 3 sources

Opinion — commentary, not a factual news event.

Simon Willison says LLMs make the old 'open source means you can fix it yourself' promise actually real. He now has AI agents clone and build repos for him, no compiling headaches required.

Simon Willison has a habit these days: several times a day he'll ask Claude or Codex to grab a GitHub repo, get it compiling, and explain how some piece of it works. Ten minutes later he checks back in. That's the whole workflow. No manual dependency wrangling, no afternoon lost to a broken build script.

This is a response to a Hacker News discussion titled 'Devtools must be open source,' and Willison's angle cuts against the usual open-source pitch. The classic argument is that open code gives users freedom to inspect and change it. But Willison points out that freedom was mostly theoretical for almost everyone, including plenty of professional programmers. Reading someone else's codebase and successfully patching it takes real time, and most people, most of the time, just didn't have it. The practical benefit of open source was less about self-reliance and more about knowing someone else, somewhere, could and would do that work for you.

Willison thinks coding agents are quietly closing that gap. The friction that used to kill curiosity, getting a project to even build locally, has dropped close to zero when you can hand that chore to an AI and walk away. He's not yet at the point of routinely rewriting the tools he depends on. But he says the path to doing that, which barely existed a year ago, is now visible in a way it simply wasn't before.

It's a small, personal observation rather than a grand prediction, and Willison frames it that way. Still, it lands on something real about how the calculus of 'is this worth my time' shifts when an agent absorbs the boring 20% of the job. The open-source promise of user-modifiable software wasn't broken by bad intentions. It was broken by the cost of attention. Cheapen that cost and the old argument starts working again.

My take

I've been saying for a year that the real open-source unlock from LLMs isn't code generation, it's code comprehension at scale, and Willison's ten-minute clone-and-explain habit is exactly that playing out. The people who kept insisting open source only matters for large institutions with dedicated engineers were always underestimating how much of 'freedom to modify' was gatekept by tedium rather than skill. Cheap agentic labor doesn't make open source obsolete, it makes the founding argument for it true for the first time in decades.

Read more about this at: Simon Willison

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.