simonw

11 ms·

1. 211.▲
   by simonw 1mo ago

   llm -m meta-ai/muse-spark-1.3 "Generate an SVG of a pelican riding a bicycle" https://tools.simonwillison.net/markdown-svg-renderer?url=ht... 4.2266 cents, 38 seconds. For comparison here's Muse
2. 212.▲
   by simonw 1mo ago

   I didn't say trained on, I said "trained for SVG output". Gemini team members have publicly stated that they have trained for SVG: https://twitter.com/sunjiao123sun_/status/202455551655137292... &gt
3. 213.▲
   by simonw 1mo ago

   Are you a frequent user of LLMs? That "copy pasting whole blocks" mental model doesn't hold up to regular usage, in my opinion.
4. 214.▲
   by simonw 1mo ago

   This thing works using WebAssembly - https://claude.com/_next/static/media/c2pa_verify_wasm_bg.53... - which is a WASM compiled version of this Rust library: https://github.com/contentauth&#x2
5. 215.▲
   by simonw 1mo ago

   Here's quite an impressive follow-up. I have a tool which knows how to render Markdown documents with embedded SVG content - I use it for the pelican test. Since this transcript has HTML in it, I decided to upgrade that tool to also re
6. 216.▲
   by simonw 1mo ago

   Hah, the pelican is in this other comment: https://news.ycombinator.com/item?id=49537553#49538217
7. 217.▲
   by simonw 1mo ago

   The speed combined with the fact that this thing is really good at HTML JavaScript is pretty exciting. Here's what I got for 1.8 cents and 13 seconds from the prompt "make me a cool thing in html": https://gisthost
8. 218.▲
   by simonw 1mo ago

   Watch the video. It's from then-Gemini-lead Jeff Dean and the video shows off an animated pelican riding a bicycle, a frog on a penny-farthing, a giraffe driving a tiny car, an ostrich on roller skates, a turtle kickflipping a skateboa
9. 219.▲
   by simonw 1mo ago

   The most interesting thing about the Gemini models is still their multi-modal support: they accept audio and video input, OpenAI and Anthropic's flagships are still image-only. Gemini Flash is also pretty cheap, so it's a great fa
10. 220.▲
    by simonw 1mo ago

    The Gemini models have openly trained for SVG output, apparently with a specialism on animals in forms of transport! https://twitter.com/JeffDean/status/2024525132266688757
11. 221.▲
    by simonw 1mo ago

    Congratulations, you're this thread's "pelicans are tiresome" comment - it's part of the Hacker News tradition at this point. (Next up is the comment saying that the labs are clearly training for the benchmark.)
12. 222.▲
    by simonw 1mo ago

    Pelicans (thinking effort high, medium, low): https://tools.simonwillison.net/markdown-svg-renderer?url=ht... - high cost 8.9742 cents Here are the 3.7 pelicans for comparison: https://tools.simonwillison.net&#x2
13. 223.▲
    by simonw 1mo ago

    That's the nature of installing software. I'd like to know OpenAI's approach to updating these things - their auto-updated is pretty aggressive already, but does that cover their vendored cached apps too?
14. 224.▲
    by simonw 1mo ago

    How do you mean? I trust OpenAI to have installed apps that at least work well enough for their intended purposes.
15. 225.▲
    by simonw 1mo ago

    The thing about $1.65T being "hidden from the balance sheets" is overblown. How do we know about that money if it's hidden? Because it isn't - it's in other public SEC documents, which the $1.65T claim also slightly
16. 226.▲
    by simonw 1mo ago

    A shill for the AI industry wouldn't spend four solid years writing about how inherently insecure their products are. https://simonwillison.net/tags/prompt-injection/
17. 227.▲
    by simonw 1mo ago

    I don't think "not being Ed Zitron's target audience" works as a useful definition of AI psychosis.
18. 228.▲
    by simonw 1mo ago

    I really, really need to pull the trigger on a dedicated site for that. Maybe I'll have Fable 5.1 build it. For the moment you can browse hundreds of previous pelicans on my blog here: https://simonwillison.net/tags&#x2
19. 229.▲
    by simonw 1mo ago

    The animation isn't down to max though - I took the SVG created by max and piped it back into high and asked that to animate it.
20. 230.▲
    by simonw 1mo ago

    To me, the biggest difference is the beak: I like the max one much more than the xhigh one. Max also made a nicer background.
21. 231.▲
    by simonw 1mo ago

    For the transcripts I shares above I'm using my https://llm.datasette.io/ tool - it has the ability to export logs to markdown: llm logs -cue > transcript.md Then I paste that into a Gist, then paste the Gist
22. 232.▲
    by simonw 1mo ago

    Are any of those predictions that have been proved right yet?
23. 233.▲
    by simonw 1mo ago

    Consider this original comment: https://news.ycombinator.com/item?id=49526069#49527546 I'll expand the section of the article that it quotes: > Although this wouldn't be in the spirit of Zitron's statement
24. 234.▲
    by simonw 1mo ago

    His "Unethical guide to surviving AI layoffs" was solid gold: https://www.youtube.com/watch?v=JgVBqcqUGE0 > Now, if your CEO has never heard the phrase Ralph Loop, oh man, you are less than 30 days away from yo
25. 235.▲
    by simonw 1mo ago

    I don't follow.
26. 236.▲
    by simonw 1mo ago

    Right, you're very much his target audience.
27. 237.▲
    by simonw 1mo ago

    They don't publish their AI spending, revenue, and profit - but they do tend to break out other non-AI segments of their companies which can demonstrate that at least part of their growth isn't relevant to AI. Page 24 of Amazon&#x
28. 238.▲
    by simonw 1mo ago

    I don't see anything wrong with it, I just think it's interesting.
29. 239.▲
    by simonw 1mo ago

    I'm not sure. Someone who hasn't installed the ChatGPT (or Codex) apps yet could confirm this by installing the apps, seeing if that ~/.cache directory exists, then try running a prompt that needs Python or Node.js or LibreOf
30. 240.▲
    by simonw 1mo ago

    (This is more of a tweet than a blog post, it really wasn't written with Hacker News in mind.)