simonw

22 ms·

601.
by simonw 3mo ago
I'm suggesting that you can't release a fix for an issue that hasn't been reported to you yet.
602.
by simonw 3mo ago
According to https://cloudsecurityalliance.org/artifacts/hugging-face-cis... the models ended up finding CyberGym solutions, which was the wrong benchmark.
603.
by simonw 3mo ago
That video suggests that RTDEV-92030 was the fix for the issue - but https://docs.jfrog.com/releases/docs/artifactory-self-manage... says that issue was resolved in a 15 July 2026 release of Artifactory, which doe
604.
by simonw 3mo ago
This appears to confirm that the packaging proxy they were using for the research machine that their agent broke out of was Artifactory, the same software they use in production. (You can tell they use it in production by asking regular Cha
605.
by simonw 3mo ago
It sounds to me like you are blaming imagined job losses on anyone who wrote about an interesting experiment that a company did.
606.
by simonw 3mo ago
What do you mean by "completely backwards"?
607.
by simonw 3mo ago
Have you really seen engineers fired for failing to spend $1,000 on tokens a day because their bosses read my piece about Strong DM? https://simonwillison.net/2026/Feb/7/software-factory/ I would hope th
608.
by simonw 3mo ago
Bun had an enormous existing test suite written in TypeScript. Getting those tests to pass against the Rust version was the key thing that enabled the project. You can go and check if the AI edited the tests yourself: look at the git histor
609.
by simonw 3mo ago
I think the prices Uber et al are paying - the list prices for the API - pay out a healthy margin to the AI labs. They're not profitable against the R&D spend, but they could be profitable if the AI companies attract more paying cu
610.
by simonw 3mo ago
You mean Strong DM? I stand by what I said about them at the time: they're an interesting case-study in extreme AI-assisted software development, and they have a bunch of ideas that are worth learning from even if you're not going
611.
by simonw 3mo ago
These distributions are excellent. Astral took over maintenance of them a while ago so technically they sit under OpenAI now. https://github.com/astral-sh/python-build-standalone If you're looking to bundle Python
612.
by simonw 3mo ago
> I've heard stories of LLMs changing tests to get all the tests to pass instead of actually fixing code. That's very easy to prevent. Don't let them edit the tests! Run the test suite against a reserved copy.
613.
by simonw 3mo ago
Presumably because they "build for [macOS, Linux, FreeBSD, Android, Windows] x [ARM64, x64]" and self hosting all of that would be time-consuming and expensive.
614.
by simonw 3mo ago
I'm trying to make fetch happen.
615.
by simonw 3mo ago
I expect everyone understood that. The obvious second effect is that if you have a leaderboard people will compete to top it, by running dumb loops or whatever. That impacts how much you can learn, and costs you more money than you wanted
616.
by simonw 3mo ago
I use the term "janky" for any time someone releases something under a supposedly open source license that doesn't comply with the OSI definition. "Modified MIT" is the perfect example of that. I'm not saying i
617.
by simonw 3mo ago
It's available for arbitrary use if you run the canary release.
618.
by simonw 3mo ago
I try not to publish articles that call out mistakes by other people just for the sake of it. I've written a fair amount about the Bun rewrite already - https://simonwillison.net/search/?q=bun%20rust
619.
by simonw 3mo ago
Good catch, thanks.
620.
by simonw 3mo ago
Yeah, tokenmaxxing was obviously stupid. Setting a limit is what happens when the adults in the room stop being stupid.
621.
by simonw 3mo ago
Kimi 2.6 had the same janky license: https://huggingface.co/moonshotai/Kimi-K2.6/blob/main/LICENS... - looks like they've been doing that at least as far back as K2. The models they released in 2025
622.
by simonw 3mo ago
One big difference I have with Ed Zitron is I look at companies suddenly getting worried that they're spending millions of dollars on tokens and think "wow those AI vendors are going to make SO MUCH MONEY". The best price for
623.
by simonw 3mo ago
Why are you resistant to adding a sentence to the article that notes that Claude Code uses the rewrite? Do you think it would weaken the article? When I said it would improve the credibility I did mean it. My instinct on reading the article
624.
by simonw 3mo ago
Yeah, I would like my taxes to go to paying for expertise to help make good governance decisions, personally.
625.
by simonw 3mo ago
In this sentence, sure: > Anthropic is dogfooding this, the machine is still ticking along, and Anthropic employees are directly involved. The problem is that your article's central claim is that there hasn't been a Bun release
626.
by simonw 3mo ago
> OpenAI nearly doubled its federal lobbying expenditure to a record $2.22mn in the first half of 2026, compared to last year, while Anthropic nearly tripled its spending to $3.53mn, according to federal disclosures. Never ceases to amaz
627.
by simonw 3mo ago
It compiles to C if you use "--backend c" Otherwise it looks like it compiles to LLVM.
628.
by simonw 3mo ago
Yeah I didn't think I'd be mistaken for being involved in the project! I deleted two words from the README.
629.
by simonw 3mo ago
No, it's a raw count. I get that there's a baseline of unsafe that's necessary, but I was under the impression that some of those unsafes were in places where they could be replaced by more idiomatic Rust at a later date.
630.
by simonw 3mo ago
One metric I thought might be interesting with respect to the Rust rewrite is the number of "unsafe" occurrences in the code. I was optimistic those would be going down over time but it looks like they've stayed pretty consta

More ›