21–30 of 203 posts
Supposing agents and their organization improve, it seems like we’re approaching a point where the cost of a piece of software will be driven down to the cost of running the hardware, and the cost of the tokens required to replicate it. The tokens were “expensive” from the minds of humans …
"To test this system, we pointed it at an ambitious goal: building a web browser from scratch." I shared my LLM predictions last week, and one of them was that by 2029 "Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise" https://simonwillison.net/2026/Jan/8/llm-predictions-for-202... and https://www.youtube.com/watch?v=lVDhQMiAbR8&t=3913s This project from Cursor is the s…
This is going to sound sarcastic, but I mean this fully: why haven't they merged that PR. The implied future here is _unreal cool_. Swarms of coding agents that can build anything, with little oversight. Long-running projects that converge on high-quality, complex projects. But the examples feel thin. Web browsers, Excel, and Windows 7 exist, and they specifically exist in the LLM's training sets. The closest to real…
Pretty much everything exists in the training sets. All non-research software is just a mishmash of various standard modules and algorithms.
But what would be the point of re-creating existing applications? It would be useful if you can produce a better version of those applications. But the point in this experiment was to produce something "from scratch" I think. Impressive yes, but is it useful?
A more practically useful task would be for Mozilla Foundation and others to ask AI to fix all bugs in their application(s). And perhaps they are trying to do that, let's wait and see.
"To test this system, we pointed it at an ambitious goal: building a web browser from scratch." I shared my LLM predictions last week, and one of them was that by 2029 "Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise" https://simonwillison.net/2026/Jan/8/llm-predictions-for-202... and https://www.youtube.com/watch?v=lVDhQMiAbR8&t=3913s This project from Cursor is the s…
2029? I have no idea why you would think this is so far off. More like Q2 2026.
> Another experiment was doing an in-place migration of Solid to React in the Cursor codebase. It took over 3 weeks with +266K/-193K edits. As we've started to test the changes, we do believe it's possible to merge this change.
In my view, this post does not go into sufficient detail or nuance to warrant any serious discussion, and the sparseness of info mostly implies failure, especially in the browser case.
It _is_ impressive that the browser repo can do _anything at all_, but if there was anything more noteworthy than that, I feel they'd go into more detail than volume metrics like 30K commits, 1M LoC. For instance, the entire capability on display could be constrained to a handful of lines that delegate to other libs.
And, it "is possible" to merge any change that avoids regressions, but the majority of our craft asks the question "Is it possible to merge _the next_ change? And the next, and the 100th?"
If they merge the MR they're walking the walk.
If they present more analysis of the browser it's worth the talk (not that useful a test if they didn't scrutinize it beyond "it renders")
Until then, it's a mountain of inscrutable agent output that manages to compile, and that contains an execution pathway which can screenshot apple.com by some undiscovered mechanism.
> While it might seem like a simple screenshot, building a browser from scratch is extremely difficult. > Another experiment was doing an in-place migration of Solid to React in the Cursor codebase. It took over 3 weeks with +266K/-193K edits. As we've started to test the changes, we do believe it's possible to merge this change. In my view, this post does not go into sufficient detail or nuance to warrant any seriou…
But is this actually true? They don't say that as far as I can tell, and it also doesn't compile for me nor their own CI it seems.
The browser it built, obviously the context window of the entire project is huge. They mention loads of parallel agents in the blog post, so I guess each agent is given a module to work on, and some tests? And then a 'manager' agent plugs this in without reading the code? Otherwise I can't see how, even with ChatGPT 5.2/Gemini 3, you could do this otherwise? In retrospect it seems an obvious approach and akin to how…
Who created those agents and gives them the tasks to work on. Who created the tests? AI, or the humans?
Supposing agents and their organization improve, it seems like we’re approaching a point where the cost of a piece of software will be driven down to the cost of running the hardware, and the cost of the tokens required to replicate it. The tokens were “expensive” from the minds of humans …
It will be driven down to the cost of having a good project and product manager effectively understanding what the customer wants, which has been the main barrier to excellent software for a good long time.
I think this situation emphasizes the importance of (something like) Agile. To produce something useful can only happen via experimentation and getting feedback from actual users, and re-iterating relentlessly.
This is going to sound sarcastic, but I mean this fully: why haven't they merged that PR. The implied future here is _unreal cool_. Swarms of coding agents that can build anything, with little oversight. Long-running projects that converge on high-quality, complex projects. But the examples feel thin. Web browsers, Excel, and Windows 7 exist, and they specifically exist in the LLM's training sets. The closest to real…
because it is absolutely impossible to review that code and there is gazillion issues there.
The only way it can get merged is YOLO and then fix issues for months in prod which kinda defeats the purpose and brings gains close to zero.