simonw

12 ms·

1. 271.▲
   by simonw 1mo ago

   It's effectively the source code for ChatGPT Work. If you don't want to see that, don't visit it. The non-vomit version of it is my (entirely human written) OP: https://simonwillison.net/2026/Aug/30&
2. 272.▲
   by simonw 1mo ago

   I expect OpenClaw was a huge influence on this, and on Claude Cowork too. OpenClaw demonstrated that there was enormous existing demand for general agent functionality, such that people would buy a whole Mac mini just to get access to thi
3. 273.▲
   by simonw 1mo ago

   I had it build me this website about medieval pelican Christian iconography: https://london-pelicans-in-her-piety.simonw.chatgpt.site/
4. 274.▲
   by simonw 1mo ago

   "It is an extraordinarily confusing and very powerful product." "I find [OpenAI's explanation] almost entirely useless" "An open question for me right now is how safe all of this stuff is." "Figurin
5. 275.▲
   by simonw 1mo ago

   OpenClaw is built on top of Pi.
6. 276.▲
   by simonw 1mo ago

   Find me a clearer explanation of what ChatGPT Work actually is. I wrote this because I got frustrated waiting for someone else to figure that out and write about it.
7. 277.▲
   by simonw 1mo ago

   That's not true of ChatGPT Work (Cloud), accessed through the mobile apps or the ChatGPT website. It IS true of ChatGPT Work (Local), accessed via the ChatGPT desktop app. I agree with you: Expecting non-devs to understand that distinc
8. 278.▲
   Wrapture(grahamdumpleton.me)

   1 points by simonw 1mo ago|0 comments
9. 279.▲
   by simonw 1mo ago

   I just updated the article to link to this site: https://codex-tool-reference.simonw.chatgpt.site/ Which I created using this prompt in a fresh Work session: > Build a site that lists every one of your tools - nearly gro
10. 280.▲
    by simonw 1mo ago

    Gemini have done exactly that. (I doubt it's because of my stupid benchmark, though!)
11. 281.▲
    by simonw 1mo ago

    The worked-example effect doesn't say that worked examples replace exercises. It says that worked examples prior to exercises beats exercises alone. There's a section on that page about faded worked examples: > "In order t
12. 282.▲
    by simonw 1mo ago

    Sharing the session would be better. What you got at the end is almost unreadable. I had to paste it into Claude to try and dig the underlying argument out. The best version I got from that (and it's still not great) was: Generatio
13. 283.▲
    by simonw 1mo ago

    Yeah. "The sequence is the whole story: overflow, then reduce, then squeeze, then stop when nothing can die." "the decoupling is complete: skill is cheap, taste is scarce, and nobody trained to spot it." "That is th
14. 284.▲
    by simonw 1mo ago

    Do you have any mechanisms in place to prevent two separate instances of OpenRun that share the same configuration from corrupting their shared S3 Litestream store? UPDATE: Looks like you answered that here already: https://news.
15. 285.▲
    by simonw 1mo ago

    I'm pretty sure that first one can be achieved using MicroPython or CPython. Your point 2 is very convincing: I hadn't realized Monty had serializable state as a core feature, that's a very compelling attribute for an agent r
16. 286.▲
    by simonw 1mo ago

    Have you heard of the worked-example effect https://en.wikipedia.org/wiki/Worked-example_effect ? It's quite a surprising result: it turns out that there are cases where seeing someone else work through a worked e
17. 287.▲
    by simonw 1mo ago

    Yeah, that's a fair point. I have plenty of older projects where I no longer understand how they work despite having written the code myself. I guess the key thing is that you need to be able to demonstrate to yourself that you underst
18. 288.▲
    by simonw 1mo ago

    If you're going to run Python code inside a WebAssembly container anyway why go with the Monty subset when you could use full CPython (or MicroPython) compiled to WASM instead?
19. 289.▲
    by simonw 1mo ago

    I don't see them doing that. From the article summary: > From October 20, applicants to the Fair Work Commission will need to disclose AI use, while a template to help AI-dependent litigants has been implemented. That sounds very fa
20. 290.▲
    by simonw 1mo ago

    Can you provide steps to reproduce so I can see one of these deterministic LLMs running myself? API based or local models.
21. 291.▲
    by simonw 1mo ago

    By "can't predict exactly where the weights will take you next" I meant with your brain. The blind chess analogy suggests you can predict, using your own thought process, the exact output of a prompt.
22. 292.▲
    by simonw 1mo ago

    I found this formal decision PDF for the Khan case: https://fwcgovau.sharepoint.com/teams/group252c29c/_layouts/... Key section: > Two days after Mr Khan filed his claim, Fair Work Commission staff sent M
23. 293.▲
    by simonw 1mo ago

    The way I see it, LLMs are a direct attack on friction designed into human processes. - Filing a lawsuit - Applying to a university - Tailoring a cover-letter for a job application - Writing to a local government body to complain about st
24. 294.▲
    by simonw 1mo ago

    I'm not completely convinced by this comparison between blind chess and prompting LLMs. In blind chess you get deterministic information about the state of the board: each mental update to your board model can be precise, and you have
25. 295.▲
    by simonw 1mo ago

    It's common for models to produce a draft of the SVG part way through their reasoning. Here's Qwen3.8-Flash-Next doing that, for example: https://tools.simonwillison.net/markdown-svg-renderer#url=ht... The open we
26. 296.▲
    by simonw 1mo ago

    > [...] Let's maybe add a helmet? It could improve riding theme, but may obscure head. Maybe a small cycling cap or helmet? The user didn't ask; can add red helmet? Might be cute. But pelican with big beak; a helmet might obscu
27. 297.▲
    by simonw 1mo ago

    If you don't know the song Welcome to the Internet by Bo Burnham, allow me to introduce you to it: https://www.youtube.com/watch?v=k1BneeJTDcU (Heck of a karakoe number if you can pull it off.)
28. 298.▲
    by simonw 1mo ago

    We go to them because that's where the friends and organizations we care about have chosen to post their content.
29. 299.▲
    by simonw 1mo ago

    A migration document like this one is how you get that ~free upgrade - it's something you can point an agent at.
30. 300.▲
    by simonw 1mo ago

    HTTPX stable hasn't had a release since December 2024, but there have been two dev releases of the 1.0 branch this month (after a gap since September 2025.) Unfortunately that 1.0 work is happening in a private repository. https:&#x2F