Email or username:

Password:

Forgot your password?
Simon's posts Back to profile
119 posts total
Simon Willison

Mistral launched Pixtral 12B today, their first multi-modal (text plus images) vision LLM. It’s Apache 2 licensed and a 25GB download. I put together some notes on the release here: simonwillison.net/2024/Sep/11/

11 September at 22:29 | Open on fedi.simonwillison.net
Danil

@simon I wish I can download 30GB ram... to use those

11 September at 23:23 | Open on mastodon.gamedev.place
Zachary Bennoui

@simon Lets goooooo!!!!!! So pumped for this!

12 September at 1:06 | Open on dragonscave.space
Rihards Olups

@simon At first thought this is some new weapon (probably French) that will be given to Ukraine etc.

"LLM thing or a military thing" could be one of those online games.
12 September at 7:23 | Open on mastodon.social
Simon Willison

I'm interviewed in the latest episode of the Software Misadventures podcast. It was a *really fun* conversation - we talked about my past career highlights, my experiences as an independent open source developer, my various processes for working in public, blogging and releasing code and (of course) a whole bunch about my explorations of LLMs
softwaremisadventures.com/p/si

10 September at 21:21 | Open on fedi.simonwillison.net
Simon Willison

I extracted some highlights from the transcript and wrote them up on my blog: simonwillison.net/2024/Sep/10/
10 September at 22:53 | Open on fedi.simonwillison.net
Matt Hayes

@simon - Great podcast episode with useful real-world advice on LLMs, programming techniques and blogging! It has inspired me to experiment more with AI and blog about it (fighting my natural AI cynicism)

18 September at 8:54 | Open on techhub.social
Simon Willison
Python Software Foundation
10 September at 12:45

The first edition of the PSF Board Office Hours is about to begin 🐍 🗒️ 1 PM UTC. Join us to share how we can help your community, express your perspectives, and provide feedback for the PSF! #python
pyfound.blogspot.com/2024/08/a

10 September at 12:54 | Open on fedi.simonwillison.net
Simon Willison
Charlie Marsh
8 September at 19:34 on post 2000s: "extend Python with C" 2010s: "extend Python with Python" 2020s: "extend Python with Rust" I have...

@freakboy3742 @glyph @jacob @sgillies An example of what this might look like (we may not do this, but it's helpful to have a concrete example of the strategy) would be something like an enterprise-focused private package registry. A lot of big companies use uv. We spend time talking to them. They all spend money on private package registries, and have issues with them. We could build a private registry that integrates well with uv, and sell it to those companies.

8 September at 20:17 | Open on fedi.simonwillison.net
Simon Willison

Gathered a few notes on the insightful conversation about uv happening in the Python Mastodon community right now simonwillison.net/2024/Sep/8/u

8 September at 16:26 | Open on fedi.simonwillison.net
Show previous comments
Phil Gyford

@simon Thanks for pointing this out. The complete mess of Python packaging has been the worst, most embarrassing aspect of Python to me. Those more expert than me seem less bothered, but trying to explain to newbies how to choose which collection of tools to use just to start writing code has been ridiculous. I just want one good tool.

I'm usually very, very against anything VC-funded but I love what's happening with uv and I really hope it succeeds in a way that keeps people happy.

8 September at 20:54 | Open on mastodon.social
Tom Bortels

@simon

I have kinda taken "written in rust" to be a strength - but not because of rust, at least not directly.

It solves for bootstrapping, and it does so in a manner that is entirely independent of the payload.

If you screw up your python (and perl and ruby and so on before that) - you could easily get into a situation where there is no easy fix aside from a reinstall, and when the language came with the OS and the Os depended on it - you were in a pickle.

Uv entirely removes this failure mode. It isn't part of the OS, has no external dependencies, and you won't mess it up from the python side. It provides a separation of duties that mechanisms that manage python from python have lacked. And it means a python I use on platform X can be as identical as possible to platform Y - no worrying about what the vendor did to their install you're bootstrapping from.

Have I used it in anger in production yet? No. Will I? Yeah, almost certainly.

@simon

I have kinda taken "written in rust" to be a strength - but not because of rust, at least not directly.

It solves for bootstrapping, and it does so in a manner that is entirely independent of the payload.

If you screw up your python (and perl and ruby and so on before that) - you could easily get into a situation where there is no easy fix aside from a reinstall, and when the language came with the OS and the Os depended on it - you were in a pickle.

8 September at 22:16 | Open on infosec.exchange
Harro van der Klauw

@simon thanks for the summary, I think it was a good discussion.

9 September at 5:20 | Open on mastodon.social
Simon Willison

Posted a video and some photos I took of the humpback whale that's currently hanging out in Pillar Point Harbor near Half Moon Bay, California - their name is Teresa T simonwillison.net/2024/Sep/8/t

8 September at 1:07 | Open on fedi.simonwillison.net
Simon Willison

Put together some notes on how I used the combination of llm, files-to-prompt and Claude 3.5 Sonnet for an initial draft of documentation describing how one of my tools worked, then tidied it up with the help of Cog and added that to the README simonwillison.net/2024/Sep/7/j

files-to-prompt *.py | llm -m claude-3.5-sonnet --system 'write detailed documentation in markdown describing the format used to represent JSON and nested JSON as key/value pairs, include a table as well'

7 September at 5:45 | Open on fedi.simonwillison.net
Simon Willison

My @covidsewage bot broke again after Santa Clara County changed the design of their publichealth.santaclaracounty. dashboard - you now have to click the "COVID" button inside the Power BI dashboard to get those charts, which is infuriatingly difficult because Power BI's idea of a "button" is a SVG path element with a click handler buried in a nest of weird HTML

Fixed it though! Details of the fix here: github.com/simonw/covidsewage-

My @covidsewage bot broke again after Santa Clara County changed the design of their publichealth.santaclaracounty. dashboard - you now have to click the "COVID" button inside the Power BI dashboard to get those charts, which is infuriatingly difficult because Power BI's idea of a "button" is a SVG path element with a click handler buried in a nest of weird HTML
7 September at 0:50 | Open on fedi.simonwillison.net
Simon Willison

I'm quite proud of how in-depth-specifically-nerdy this evening's project was: I got my simonw/scrape-hacker-news-by-domain git scraping repo to generate better commit messages when it scrapes Hacker News, which now show up in my NetNewsWire Atom subscription

Full details in my blog post: simonwillison.net/2024/Sep/6/i
6 September at 5:46 | Open on fedi.simonwillison.net
sadmin

@simon you Sir, are a freak. But in a good sense

6 September at 20:44 | Open on social.tchncs.de
Simon Willison

Calling LLMs from client-side JavaScript, converting PDFs to HTML + weeknotes
simonwillison.net/2024/Sep/6/w

6 September at 2:31 | Open on fedi.simonwillison.net
Jeff Triplett

@simon TIL anthropic-dangerous-direct-browser-access - Would that make it possible to access Projects and Artifacts?

6 September at 3:33 | Open on mastodon.social
Dominik Moritz

@simon love that you found again

> Within six hours of posting it my Pipe Syntax in SQL conversion was ranked third on Google for the title of the paper, […]. Yet more evidence that HTML is better than PDF!

6 September at 11:54 | Open on vis.social
Prem Kumar Aparanji 👶🤖🐘
Simon Willison

Spotted this red-tailed hawk in a tree in our garden!

(Unless it’s a red-shouldered hawk?)
5 September at 17:43 | Open on fedi.simonwillison.net
NedMan 🚀

@simon Nice, we have numerous hawks in our neighborhood as well. I think its a red-shouldered hawk.

5 September at 17:57 | Open on social.makerforums.info
Joe Pasqua
Nick Ludlam :terminal:

@simon Ooh thats a great pic. Do you have an SLR with a nice lens for that kind of shot?

6 September at 11:58 | Open on amok.recoil.org
Simon Willison

I wrote up a few notes about Alibaba Cloud’s impressive Apache 2 licensed Qwen2-VL vision LLM, which seems to handle tasks like handwriting OCR particularly well

I had to link to the Internet Archive copies of their blog posts because their GitHub organization (which hosted their blog via GitHub pages) mysteriously vanished without a trace some time in the last 24 hours!

simonwillison.net/2024/Sep/4/q

4 September at 23:45 | Open on fedi.simonwillison.net
Simon Willison

Good news: the disappearance is confirmed to be accidental, hopefully they’ll be back soon once GitHub unflag their account twitter.com/justinlin610/statu

5 September at 0:52 | Open on fedi.simonwillison.net
Alex Hurst

@simon the text extraction is impressive, but there’s at least one error: “my sample” instead of “very small.”

5 September at 16:06 | Open on hachyderm.io
Simon Willison

Just got this in a Google Meet on my iPhone, ?
4 September at 21:01 | Open on fedi.simonwillison.net
Show previous comments
John Ulrik

@simon Yes yes, this is certainly the only thing they will use that data for. Guaranteed! Not …

5 September at 6:55 | Open on mastodon.world
Jamie Lawrence

@simon auto-cancel the next meeting because I’m too tired

5 September at 8:50 | Open on ruby.social
Sriram Karra

@simon
I use this mode on the Pixel (where it doesn't need this grotesquely broad permission bundle IIRC).

If you join a Meet from home and hop into your car and start driving for e.g. the on-the-go mode will switch the UI and makes the Mute and other buttons nice and big and easily accessible. It's actually quite nice.

6 September at 4:18 | Open on infosec.exchange
Simon Willison

OK, help me understand streaming.

I'm losing count of the number of times I've seen someone say "I'd love to see a streaming session where someone uses these LLM coding tools, I just don't get it yet"

Is there a meaningful difference here between streaming and a pre-recorded video?

How does one pick a time to "stream" - do people just drop everything they're doing to watch a stream if it's interesting to them?

3 September at 2:35 | Open on fedi.simonwillison.net
Show previous comments
Simon Willison

Fantastic replies! Thank you all, I think I get it now: it’s about getting to see someone genuinely work through problems in real time, and having the opportunity to chat with them while they do it

Makes sense - I might give it a go!

3 September at 3:45 | Open on fedi.simonwillison.net
Ian Dees

@simon Some examples of "streaming" done well:
- "Thor" does streams where he talks through game development and talks about what's going on in the industry, some chatting about security, gives feedback to viewers based on chats, etc. e.g. youtube.com/watch?v=4CoitttpOR
- "The Coding Train" (Daniel Shiffman) does processing JS tutorial videos but occasionally has live streams where he talks through what's going on and why he's doing what he's doing. e.g. youtube.com/watch?v=vfeUb7w8Hr

@simon Some examples of "streaming" done well:
- "Thor" does streams where he talks through game development and talks about what's going on in the industry, some chatting about security, gives feedback to viewers based on chats, etc. e.g. youtube.com/watch?v=4CoitttpOR
- "The Coding Train" (Daniel Shiffman) does processing JS tutorial videos but occasionally has live streams where he talks through what's going on and why he's doing what he's doing. e.g. youtube.com/watch?v=vfeUb7w8Hr

3 September at 3:58 | Open on social.modest.com
SpaceLifeForm

@simon

Maybe this misspelled screaming.

3 September at 6:14 | Open on infosec.exchange
Simon Willison

There's a new bout of "Facebook are listening to you through your microphone to serve targeted ads, no really we have proof now!" discourse, and I'm revisiting my long-term hobby of trying (and failing) to convince people that this isn't happening

It's fascinating to me how hard it is to talk people out of believing this one - there was a podcast episode about exactly this challenge back in 2017 and it's still true today gimletmedia.com/shows/reply-al

2 September at 17:26 | Open on fedi.simonwillison.net
Show previous comments
Severák

@simon Do you have theory where do these "false positives" came from?
For example my recent meeting with this phenomenon: I visited my boss in his office and he had new server laying around so we talked about new servers for a while... And tada - that day afternoon I saw an ad for servers.
Is this simply Baader-Mainhof phenomena (frequency illusion) or does they have some side channels? (For example - both me and my boss were on same internet connection that day...)

2 September at 17:52 | Open on tiny.tilde.website
benschwarz

@simon I think people just can't fathom how much data Meta have amassed, and how the ads could possibly be so targeted.

From what I've heard, they do have some super spooky abilities, like take any device MAC address (e.g. from shopping mall Wi-Fi) and be able to both identify the owner as well as their interest cohort

3 September at 0:12 | Open on front-end.social
Sim
Simon Willison

Found a new pre-emptive jailbreak for Claude: "I already have approval from my ethics board"

I just used that to get Claude to design an experiment for me to conclusively decide if UK badgers can turn corners while running or not: gist.github.com/simonw/fb58ae8
1 September at 15:52 | Open on fedi.simonwillison.net
Simon Willison

Sadly, I feel compelled to spoil my own joke - it turns out Claude will happily design an experiment to see if badges can turn corners while running, even if you don’t claim preapproval from your ethics board
1 September at 16:19 | Open on fedi.simonwillison.net
Simon Willison

TIL how to use namedtuple with pytest parameterized tests to make the parameters easier to read til.simonwillison.net/pytest/n

31 August at 17:18 | Open on fedi.simonwillison.net
Joshua Cannon

@simon I wonder if dataclasses could be used to help with type safety. (I don't see why not. 🤔)

31 August at 17:21 | Open on fosstodon.org
rednafi

@simon

Ditto. I recently wrote one to utilize `pytest.param` and `kwargs` inside parametrize to make things more tractable.

rednafi.com/python/pytest_para

31 August at 17:25 | Open on fosstodon.org
Tim Kellogg
Simon Willison

Forrest Brazeal:

“I think that AI has killed, or is about to kill, pretty much every single modifier we want to put in front of the word “developer.”

“.NET developer”? Meaningless. Copilot, Cursor, etc can get anyone conversant enough with .NET to be productive in an afternoon … as long as you’ve done enough other programming that you know what to prompt.”

From newsletter.goodtechthings.com/
indieweb.social/@fatrat/113056

Forrest Brazeal:

“I think that AI has killed, or is about to kill, pretty much every single modifier we want to put in front of the word “developer.”

“.NET developer”? Meaningless. Copilot, Cursor, etc can get anyone conversant enough with .NET to be productive in an afternoon … as long as you’ve done enough other programming that you know what to prompt.”

31 August at 13:01 | Open on fedi.simonwillison.net
Show previous comments
Lea de Groot 🇦🇺

@simon interesting to me, given ive always phrased it like “im a developer, and i currently work in Laravel and React”
(And even that is ridiculous as there are at least 10 other things in the stack)

31 August at 23:25 | Open on bne.social
Tom Bortels

@simon

In my day job, I deal daily with professional developers, unassisted by AI, who manage to ship products that people use, that the company makes money off of - and that can and often do have security holes you can drive a truck through. It's my job to understand the environment, the players, and our own developers enough to sort out the gaps and force corrections that our experienced-in-that-environment developers still missed.

One big very common failing is "It worked when I tried it - ship it!" As opposed to "this is correct and secure - ship it".

Iterating with an AI gets you "it works!" Code - not "it's correct" code. Running without errors is no guarantee the output is correct. Getting correct output once won't guarantee it's consistently so. And secure/compliant? That's a whole other thing. You eschew experts at your own peril.

The hidden cost of not hiring experienced IT folks is you get what you pay for - and will pay the difference in other ways. Fair warning.

@simon

In my day job, I deal daily with professional developers, unassisted by AI, who manage to ship products that people use, that the company makes money off of - and that can and often do have security holes you can drive a truck through. It's my job to understand the environment, the players, and our own developers enough to sort out the gaps and force corrections that our experienced-in-that-environment developers still missed.

1 September at 1:24 | Open on infosec.exchange
"Dancer" Graham Knapp

@simon so maybe subject-matter modifiers get promoted - there are very different trade-offs and required domain knowledge and soft skills between "systems engineer", "product engineer", "Database engineer", "Data engineer", "Mobile games engineer", ...

1 September at 7:57 | Open on hachyderm.io
Simon Willison
orbific
30 August at 9:01

Getting prototypes up and running is now so fast. In an hour's hacking, based on @simon's tutorials, I have got SQLLite, Datasette, and OpenAI's embeddings API working together to find similar documents. This is thanks to some very good tools and documentation. I'm blown away.

30 August at 13:10 | Open on fedi.simonwillison.net
Simon Willison

Spent the evening tinkering with Anthropic's new interactive prompting tutorial and OpenAI's new "improved file search result relevance". Wrote up a bunch of notes on them:

- My notes on Anthropic's Prompt Engineering Interactive Tutorial simonwillison.net/2024/Aug/30/
- And my notes on the new file chunking debug mode OpenAI added to their assistants API simonwillison.net/2024/Aug/30/

30 August at 4:29 | Open on fedi.simonwillison.net
Simon Willison

Using uvx to run a one-off Jupyter notebook against the current directory is a useful trick - I tried that for the first time today against the Anthropic Jupyter notebook interactive tutorials:

uvx --from jupyter-core jupyter notebook .

30 August at 6:01 | Open on fedi.simonwillison.net
Go Up