@simon (year might be wrong on that post)

@simon This site is awesome!

... just spotted they have a photo of us on their Instagram! https://www.instagram.com/p/DAZRx-QSVZK/

You can try out the vision models on https://lmarena.ai/ by navigating to "Direct Chat" and then selecting those models from the dropdown menu

maybe I'll try to figure out how to make that thing into a little wee electron app.

@simon So reading this, it's basically Django Apps without the manual fiddling to integrate?

New in DJP 0.2: @carlton suggested it would be useful if plugins could say “this middleware needs to be inserted directly before/after this other middleware” - that’s now supported:

djp.Position("my_plugin.middleware.MyPluginMiddleware", before="django.middleware.common.CommonMiddleware")

https://github.com/simonw/djp/releases/tag/0.2

The getting started fits in a single image

@simon @radiac Oh that's awesome! Will have to kick the tires for the next project in that scope I try. Flask has been my default but this looks great.

@simon @radiac I came across this a few weeks back. Really impressive package for throwing together ideas with the features to then expand into a fully fledged project if you need to.

Jacob’s talk is exploring what the DSF could do if revenue quadrupled from $250,000/year to $1,000,000/year

@simon 10 years later, this is basically my job now.

@simon I do find it incredible that I don't find it incredible that I can take a crappy photo of a bird and very quickly get the bird identified (using pre-AI-hype google image search for example)

@simon Great post, Simon.

Do you have any idea why all 3 providers use POST and not GET that would work with the EventSource API?

@simon Nice.

Little note: on a recent curl, you can POST JSON with `curl --json <string>, saving the header setting.
Also, `--no-buffer` should no longer be necessary.

Update: `--no-buffer` always `fflush()`es the output in curl. So it might still be beneficiary.

Updated my TIL with example JavaScript code for streaming events from a fetch() POST API (using an async iterator function) https://til.simonwillison.net/llms/streaming-llm-apis#user-content-bonus--2-processing-streaming-events-in-javascript-with-fetch

Pesto is a very big penguin https://apnews.com/article/pesto-penguin-australia-huge-chick-aquarium-45ccd55631d14a3f986f932ffbfcdc7a

"If you go to a meeting with five people, and you’ve got a working prototype, the conversation will be so much more informed than if you go in with an idea and a whiteboard sketch.

I’ve always been a prototyper, but I feel like the speed at which I can prototype things in the past 12 months has gone up by an order of magnitude.

[...] I can tap a thing into my phone, and 30 seconds later, I’ve got a user interface in Claude Artifacts that illustrates the idea that I’m trying to explore."

@simon I found this episode quite entertaining and informative, thanks

#genAI #LLMs

@simon That was really interesting! I love this kind of behind-the-scenes look at critical institutional structures we rarely think about

This is one of my favorite genres of blog post: the article I wish I’d been able to read myself

Oh lord of the rings, this resonates so much with my experience of a former board member of @gnome

Many insightful takes in this post, I definitely think GNOME and the PSF should try to work together more

@simon @ThePSF

@simon Not many, probably. Should have won an Academy Award for animation, imho. I was literally watching with my mouth open when I saw it at the cinema.

@simon fully co-signed +“self love” by coi leray and metro boomin is an absolutely perfect soundtrack to that opening

@simon That whole movie is chef's kiss.

@simon Most of my PDF creation is solving the “how do I save the current content of a webpage as a single file and not just an image” problem.

SingleFile <https://github.com/gildas-lormeau/SingleFile> and/or MHTML <https://en.wikipedia.org/wiki/MHTML> should be more universal than they are.

@simon or from an old ssl invalid website

One of my favourite documents in this genre remains "THE ELEVEN LAWS OF SHOWRUNNING", about how to run production of a TV show

It's unintentionally a fantastic guide to being an engineering manager, or startup founder, or indeed any role where you need a bunch of different creative people to understand and work together to bring about a creative vision:

https://simonwillison.net/2019/Feb/19/eleven-laws-showrunning/

@simon I always find things from the entertainment industry interesting because they have a very different definition of done than we do. Nobody is expecting changes to a video which is a month old, much less 10 years old, and that’s such a different concept than most software projects have.

@simon I found this an interesting read from a management point of view. I don’t launch people into space but have found value in the ideas around planning and testing. https://ntrs.nasa.gov/citations/19720005243

@simon Maybe the issue isn't the fence-post-movers (people saying "this isn't intelligent yet"), but the fence-post-planters: i.e., those boldly offering a definition of intelligence, usually one that depends on performing a given task.

The "is it AI yet?" game may be decades old now, but the game of defining measures of intelligence that turn out to be naive and/or racist is much older, as Stephen J. Gould describes in The Mismeasure of Man.

@simon This reflects the difficulty of defining "intelligence". When a previously-unsolved problem is solved via a method that is clearly ad hoc, the field naturally moves the goal posts. Problem-by-problem, we discover what is easy and what is hard for computers.

@simon I remember when the idea was the computer couldn't learn. That one never made sense to me at all, even on simple computers you can write a program that seems to learn as it goes.

Being introduced to computing by terminal and then PC, I didn't understand that up to that time, most programmers had probably been working on batch processing of data, having put the human "computers" out of work.

And a new release of LLM that adds support for those models - for anyone lucky enough to have a tier 5 OpenAI API account at least https://llm.datasette.io/en/stable/changelog.html#v0-16

@simon o1 gets my unbirthday question which none of OpenAI's other models have: "Do all people have the same number of unbirthdays?"

https://chatgpt.com/share/66e3652c-7ae0-8002-94dc-263f6c70422e