@simon Small typo in your blog posts `deno juptyer install` command.

@simon I've been using Quarto to do EDA with observable in any case, but the full jupyter experience will be nice

@simon PDF gets more trust. Format it like a scientific paper and its truth! Page layout cosplay infuriates me.

@simon it's hard to make everyone happy. It's easy to fix though: the source is latex. @kfogel can you help?

@simon @osi Corporate / enterprise love a PDF accompanied with a webinar that you get when you give away your corporate contact info – pretty much the opposite of what non-corporate wants

@simon FYI @yossarian

... looks like I was having trouble here because I had Homebrew's developer mode turned on - if you don't have that mode turned on "brew update" should work correctly and the new attestation feature isn't enabled https://til.simonwillison.net/homebrew/no-verify-attestations#user-content-developer-mode

I'm trying out this filter mechanism, since they all seem to include links to the same site despite coming from different accounts on different Mastodon servers

@simon @mastohost If you're running 4.3.0, there is an option to control notifications from people you don't follow, new accounts, and people not following you.

@simon The instance I'm on silences and defederates from quite a few other instances out there; maybe that could be an option for you, too?

https://meta.chaos.social/federation

It sounds a bit like these mail spam ban lists (spamhaus I believe?) from the 90s. Not sure if every instance needs to start from scratch.

@jacob This is really interesting. Do you know the name of the plot for the budget visualization? I'd like to make a similar one; doesn't look hard to draw but maybe there are tools?

@jacob This is a great (and aesthetically pleasing!) walk through of the practicalities of OSS fundraising! Wow!

@freakboy3742 @ThePSF that's fantastic congratulations

@freakboy3742 @ThePSF Congratulations to you and Malcolm for this amazing achievement!

@freakboy3742 @ThePSF Congratulations and thank you for pursuing your ambitious goal that will benefit the entire Python community 👏

Today’s Rails 8 release has a strong SQLite flavor to it as well https://simonwillison.net/2024/Oct/7/whats-new-in-ruby-on-rails-8/

@simon That callout about URL.revokeObjectURL() amused me because in scanning the code my PR trained eyes immediately went looking for it after the create Blob. It’s an easy to overlook performance need when working with Blobs in the browser. I’ve found so many huge memory leaks from unpaired createBlob/revokeObjectURL.

One of the reasons I’m hoping the “Disposable” pattern makes into TC-39 Stage 4 soon: https://github.com/tc39/proposal-explicit-resource-management

Here's a related tool that can turn a JPEG or PNG into a vector SVG, using the https://github.com/jankovicsandras/imagetracerjs library by András Jankovics

https://tools.simonwillison.net/image-to-svg

Claude transcript for that one is https://gist.github.com/simonw/d2e724c357786371d7cc4b5b5bb87ed0

@simon in the past two days I’ve built two WASM based in-browser tools - one with ChatGPT 4o and the other with Claude 3.5. The ChatGPT tool took maybe 2 hours with revisions but the Claude generated tool took 5 minutes and exactly 2 revisions. And the best part is, I don’t speak Rust. Getting sidetracked to build a needed tool is now just a quick pull over into a rest stop instead of the 200 mile detour it used to be. It’s fantastic.

Thanks everyone! Blasted past 50, I’m still seeing this screen for some reason though

@simon AFAIK you can livestream from a mobile even with less subscribers by simply using the browser (in "desktop mode") instead of the app.

@simon

Done!

What’s weird about this is that in the past I’ve tried it with LLM text summaries of my drafts and found very little value in the result - for some reason I find the NotebookLM audio conversation format much more interesting in helping me think critically about what I’ve written

@simon Can you like, get the model to like, stop using the word like so much? 😊

@simon I dropped one of my longer blog posts into Notebook LM and had it generate a podcast and the discussion was interesting. The odd thing is it mentioned framings of things that weren't in what I wrote, I had to go back and read my post from 8 to 10 years ago and it wasn't there. Later in the discussion it got the framing right and that nuance was something intended as the usual common framing wasn’t quite correct.

It is helpful if more mainstream audiences are the audience.

@simon pair programming with a teenager interrupting your thoughts with inane comments? But also helping occasionally. Just those autocompleted comments really break my flow.

@simon For me, it’s been like having an over-anxious friend trying as hard as possible to be helpful.
Helpful, yes, but upon examination their answers are invariably in need of refinement.

@simon I always understood the one driving the keyboard wasn't the one coding: the coder had to verbalise what they wanted, but not having to type would allow them to think more.

Which seems the other way around to an LLM right now. But it does make me wonder whether a fast enough speech processor and enough IDE context could allow you lead an LLM like that...

@simon your clarity of thought is always appreciated

@simon I like this position, but I can’t trust anything Sama says about AGI. He’s one of the chief culprits churning up noise around it. He just knows how to code switch.

@simon Awesome service. Thanks a lot Simon. Also great commentary.

Small improvement for next time. Don’t refresh when text is selected or restore selection. Copying text (mobile safari) is a bit tricky.

Upgraded my live blog - you can now opt to have the most recent updates show up at the top of the page. I have pretty comprehensive notes from several sessions now, including the deep dive into the new Realtime API. https://simonwillison.net/2024/Oct/1/openai-devday-2024-live-blog/

@simon interesting they are releasing speech to text models. I wonder will anyone be releasing improved text to speech models anytime soon?

@simon on an Intel-silicon Mac, or only the Apple sand?