Simon Willison(@simonw)

Pelicans and a bunch of notes on the pricing of the new Claude Haiku 5.5 https://t.co/F3Sob8WvsG

Simon Willison (@simonw) — Wed, 07 Oct 2026 20:58:38 GMT

Pelicans and a bunch of notes on the pricing of the new Claude Haiku 5.5 simonwillison.net/2026/Oct/7/cla…

💬76🔄13❤️216👀28191📊91

I released an LLM plugin for sending prompts to OpenAI's new Jev-clone decision model that's based o...

Simon Willison (@simonw) — Wed, 07 Oct 2026 00:32:42 GMT

I released an LLM plugin for sending prompts to OpenAI's new Jev-clone decision model that's based on GPT-6 Luna simonwillison.net/2026/Oct/6/llm…

💬80🔄9❤️254👀29609📊117

One thing I'd LOVE to understand is how Claude "knows" what the music from Secret of Monkey Island s...

Simon Willison (@simonw) — Tue, 06 Oct 2026 23:33:28 GMT

One thing I'd LOVE to understand is how Claude "knows" what the music from Secret of Monkey Island sounds like

Maybe Anthropic crawled a whole bunch of online music discussion forums with text-based renditions of classic game music?

💬13🔄0❤️29👀6235📊14

Here's a benchmark that uses Lilypond https://t.co/6yWiT7tG9e

Simon Willison (@simonw) — Tue, 06 Oct 2026 17:00:55 GMT

Here's a benchmark that uses Lilypond x.com/aug5thmusic/st…

Auggie@aug5thmusic

GPT-6 Astra has the best result yet on the Bach Benchmark. Its chorale contains no voice-leading errors, and its harmonic palette is sophisticated enough to include a Neapolitan sixth chord. More importantly, it is the first model to ever write passing tones on this benchmark, a significant leap in musical understanding.

Extra High effort. Prompt below:
“In LilyPond (version 2.24), write a 4-part chorale in the style of Bach, 3/4 time, G minor. Use two staves -- soprano and alto on the top staff, tenor and bass on the bottom staff. Respond with only the code block.”

Your browser does not support the video tag.

💬8🔄2❤️32👀13958📊10

Anyone know when the models started being able to compose competent music? I had Opus 5.5 "write som...

Simon Willison (@simonw) — Tue, 06 Oct 2026 15:40:03 GMT

Anyone know when the models started being able to compose competent music? I had Opus 5.5 "write some computer game music for me" and the results were much better than I had expected
simonwillison.net/2026/Oct/6/scr…

💬131🔄17❤️308👀36772📊163

Mistral can pelican now! https://t.co/Mvtss5e5ay

Simon Willison (@simonw) — Tue, 06 Oct 2026 14:35:18 GMT

Mistral can pelican now! tools.simonwillison.net/markdown-svg-r…

Mistral AI@MistralAI

Meet Mistral Large 4, aka Le Chonk. 

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

Your browser does not support the video tag.

💬81🔄28❤️755👀30772📊111

I re-ran an experiment @colin_fraser ran against GPT-4o a while back to see how good it was at addin...

Simon Willison (@simonw) — Mon, 05 Oct 2026 17:19:20 GMT

I re-ran an experiment @colin_fraser ran against GPT-4o a while back to see how good it was at adding long numbers, only this time I tried Qwen 3.8 27B running locally in both reasoning and non-reasoning modes simonwillison.net/2026/Oct/4/qwe…

💬115🔄30❤️473👀50937📊167

Just published this post about how we’re going to need default hard budget caps on pretty much every...

Simon Willison (@simonw) — Sat, 03 Oct 2026 23:35:55 GMT

Just published this post about how we’re going to need default hard budget caps on pretty much everything simonwillison.net/2026/Oct/3/def…

💬255🔄46❤️918👀121266📊391

Now that we've had a few days with it, how are people differentiating between Dot and regular ChatGP...

Simon Willison (@simonw) — Sat, 03 Oct 2026 21:43:53 GMT

Now that we've had a few days with it, how are people differentiating between Dot and regular ChatGPT?

I'm having trouble deciding when I should prompt Dot vs using ChatGPT - my Dot seems to afford a single conversation, but I like controlling my context across multiple threads

💬180🔄3❤️465👀87177📊202

I'm at OpenAI's DevDay event in San Francisco today - as I have for the past three DevDay events, I'...

Simon Willison (@simonw) — Tue, 29 Sep 2026 16:54:44 GMT

I'm at OpenAI's DevDay event in San Francisco today - as I have for the past three DevDay events, I'm running a live blog where I'll be posting updates during the keynote, which starts in five minutes simonwillison.net/2026/Sep/29/op…

💬85🔄35❤️525👀126312📊157

More on Sonnet 5.5 https://t.co/efd6KqvkqU

Simon Willison (@simonw) — Tue, 29 Sep 2026 03:24:37 GMT

More on Sonnet 5.5 simonwillison.net/2026/Sep/28/cl…

💬15🔄3❤️52👀14339📊19

The most important thing about Sonnet 5.5 is that it's now the model that powers the free tier on ht...

Simon Willison (@simonw) — Mon, 28 Sep 2026 21:18:41 GMT

The most important thing about Sonnet 5.5 is that it's now the model that powers the free tier on claude.ai - so all of this stuff can be done by free users

ChatGPT's free tier is still GPT-5.6 Luna, which is a lot less capable

Claude@claudeai

A thread of early experiments with Claude Sonnet 5.5.

A fall foliage simulator by @_re_pete, made with Sonnet 5 vs Sonnet 5.5.

Your browser does not support the video tag.

💬188🔄81❤️1803👀267746📊286

@WeAreDevs (The full talk also covers New Zealand’s Kākāpō parrot breeding season, probably the best...

Simon Willison (@simonw) — Mon, 28 Sep 2026 17:56:46 GMT

@WeAreDevs (The full talk also covers New Zealand’s Kākāpō parrot breeding season, probably the best news of 2026)

💬1🔄1❤️21👀5302📊3

@WeAreDevs And my closing thought on how this stuff impacts our lives as software engineers, with a ...

Simon Willison (@simonw) — Mon, 28 Sep 2026 17:55:57 GMT

@WeAreDevs And my closing thought on how this stuff impacts our lives as software engineers, with a quote from cycling champion Greg LeMond

💬8🔄1❤️51👀7323📊12

@WeAreDevs Here's how I define "Fable class models" - first Claude Fable 5, now Claude Opus 5.5 and ...

Simon Willison (@simonw) — Mon, 28 Sep 2026 17:53:47 GMT

@WeAreDevs Here's how I define "Fable class models" - first Claude Fable 5, now Claude Opus 5.5 and GPT-Astra 6 and maybe GPT-5.6 Sol as well

💬26🔄3❤️74👀12577📊31

I've published detailed notes and an annotated transcript to accompany the video of the keynote I ga...

Simon Willison (@simonw) — Mon, 28 Sep 2026 17:51:39 GMT

I've published detailed notes and an annotated transcript to accompany the video of the keynote I gave at @WeAreDevs World Congress North America in San Jose on Friday - here's my rundown of everything that's happened with LLMs and agents in 2026 so far simonwillison.net/2026/Sep/27/20…

💬64🔄90❤️713👀87207📊262

Prompt and transcripts here https://t.co/QI1dIenQsW - or you can load the HTML page Claude built to ...

Simon Willison (@simonw) — Sun, 27 Sep 2026 00:19:59 GMT

Prompt and transcripts here simonwillison.net/2026/Sep/26/ka… - or you can load the HTML page Claude built to see the interactive version before it became a video tools.simonwillison.net/kakapo-party

💬16🔄3❤️25👀13332📊16

I included a few references to this year's record-breaking Kākāpō breeding season in a talk I gave y...

Simon Willison (@simonw) — Sun, 27 Sep 2026 00:18:13 GMT

I included a few references to this year's record-breaking Kākāpō breeding season in a talk I gave yesterday, and since Claude Opus 5.5 is surprisingly capable at pixel art animation I had it create this celebratory video for my closing slide

Your browser does not support the video tag.

💬75🔄6❤️151👀29117📊78

@AndyMasley OK maybe I should play some new games

Simon Willison (@simonw) — Sat, 26 Sep 2026 15:13:26 GMT

@AndyMasley OK maybe I should play some new games

💬12🔄0❤️50👀53190📊15

"The last year in the technology industry has felt like 100 years all happening at once. Our industr...

Simon Willison (@simonw) — Sat, 26 Sep 2026 03:30:59 GMT

"The last year in the technology industry has felt like 100 years all happening at once. Our industry is destabilized in a way nobody’s experienced since the advent of the personal computer. Every limit AI runs up against collapses within a month. Everything we do with frontier models today, in a few years we’ll be doing instantaneously and for free."

Thomas H. Ptacek@tqbf

Today I'm parting ways with Fly.io. It's been a privilege. I'm off to tilt at a big old windmill with Kurt, again. 

sockpuppet.org/blog/2026/09/2…

💬92🔄76❤️1033👀116561📊189

I'm on stage for the keynote in ten minutes time!

Simon Willison (@simonw) — Sat, 26 Sep 2026 00:21:55 GMT

I'm on stage for the keynote in ten minutes time!

WeAreDevelopers@WeAreDevs

Legendary programmer Simon Willison says he's never so much change happen so quickly as he has in 2026, and there's no sign of it stopping.

🎤 Don't miss his closing Keynote, Friday (Sept 25) on the Main Stage at 5.30pm!

Your browser does not support the video tag.

💬27🔄10❤️364👀59746📊51

The more time I spend working with coding agents, the more convinced I am that they make software en...

Simon Willison (@simonw) — Fri, 25 Sep 2026 01:02:12 GMT

The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder

We can do amazing things with them, but unlocking their full potential requires extraordinary discipline and knowledge

Gergely Orosz@GergelyOrosz

Bury your head in the ground at your own risk. I aim to not jump on any hype trains, but since Opus 4.6 and GPT-5.2 + the harnesses it was clear that these things can write code nearly as good as I can in my best language; better in other languages.

But we won't see non-devs push production code for a long time; possibly forever, if you ask me. Software still needs to be built in robust ways, and it's our profession to do this well, and use the new tools we have, which create new and interesting (+tough!) challenges

💬377🔄388❤️4482👀622076📊786

I just had a voice conversation with the Datasette backup of my blog via datasette-mcp, which you ca...

Simon Willison (@simonw) — Thu, 24 Sep 2026 00:22:32 GMT

I just had a voice conversation with the Datasette backup of my blog via datasette-mcp, which you can now talk to using the ChatGPT iPhone app

Vaibhav (VB) Srivastav@reach_vb

Massive QoL update: ChatGPT Voice can now use plugins AND run on GPT-6 Astra, Sol & Luna, directly in ChatGPT Work on web and mobile.

Your email, calendar, Slack + docs, decks, sites and spreadsheets, just by talking.

Rolling out now, enjoy!!

💬107🔄12❤️176👀38536📊89

Seems to cost less than 1 cent per minute of generated audio for Flash, and Flash-Lite is even cheap...

Simon Willison (@simonw) — Wed, 23 Sep 2026 20:45:37 GMT

Seems to cost less than 1 cent per minute of generated audio for Flash, and Flash-Lite is even cheaper than that

💬6🔄0❤️44👀10622📊10

The new Gemini 3.8 TTS models are super-cheap and can generate conversations between multiple voices...

Simon Willison (@simonw) — Wed, 23 Sep 2026 20:45:19 GMT

The new Gemini 3.8 TTS models are super-cheap and can generate conversations between multiple voices (from 2,000+, or you can clone your own) - I built a little UI for it, then had Claude knock up a script where two pelicans debate moving to Pacifica Pier simonwillison.net/2026/Sep/23/ge…

💬91🔄30❤️681👀46647📊210

Big model release today - I wrote about Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna - plus comparison...

Simon Willison (@simonw) — Tue, 22 Sep 2026 23:50:29 GMT

Big model release today - I wrote about Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna - plus comparison grids of pelicans by the different model families at different reasoning levels simonwillison.net/2026/Sep/22/op…

💬114🔄41❤️603👀70253📊200

GPT-6 Luna is half the price of 5.6 Luna, which was already an astonishingly cheap model given how c...

Simon Willison (@simonw) — Tue, 22 Sep 2026 20:07:37 GMT

GPT-6 Luna is half the price of 5.6 Luna, which was already an astonishingly cheap model given how capable it is

Luna is my favorite model for building product features thanks to its cost (and speed)

OpenAI@OpenAI

Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.

GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.

We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.

Your browser does not support the video tag.

💬143🔄58❤️1642👀121430📊242

Put together some notes on Jev and the new category of system one aka decision models https://t.co/E...

Simon Willison (@simonw) — Mon, 21 Sep 2026 23:16:26 GMT

Put together some notes on Jev and the new category of system one aka decision models simonwillison.net/2026/Sep/21/je…

💬155🔄93❤️1265👀93895📊451

Skeptical geneticist: "pfft, it's just frog DNA. And they deliberately let them eat people for the m...

Simon Willison (@simonw) — Sat, 19 Sep 2026 14:10:57 GMT

Skeptical geneticist: "pfft, it's just frog DNA. And they deliberately let them eat people for the marketing."

💬3🔄3❤️73👀12554📊10

Dennis Nedry works for Irregular

Simon Willison (@simonw) — Sat, 19 Sep 2026 04:21:09 GMT

Dennis Nedry works for Irregular

💬14🔄1❤️59👀25524📊17

Being a computer scientist who refuses to find anything about LLMs interesting right now is a bit li...

Simon Willison (@simonw) — Fri, 18 Sep 2026 19:19:11 GMT

Being a computer scientist who refuses to find anything about LLMs interesting right now is a bit like being a geneticist who refuses to find anything interesting about the recently opened Jurassic Park

💬230🔄186❤️3960👀163416📊608

This new Claude Code mods system looks pretty interesting - more example mods here: https://t.co/muU...

Simon Willison (@simonw) — Fri, 18 Sep 2026 19:14:07 GMT

This new Claude Code mods system looks pretty interesting - more example mods here: github.com/anthropics/cla…

Thariq@trq212

AGENTS.md support is built off of Claude Code mods, our upcoming way to customize the Claude Code harness.

This is a built-in mod, but you’ll be able to build custom versions of project instructions yourself as you’d like too.

💬11🔄3❤️21👀13704📊11

Finally! This is great, now I can stop dropping CLAUDEmd files which just contain "@AGENTSmd"

Simon Willison (@simonw) — Fri, 18 Sep 2026 19:06:20 GMT

Finally! This is great, now I can stop dropping CLAUDEmd files which just contain "@AGENTSmd"

Thariq@trq212

We're adding support for AGENTS.md to Claude Code. 

Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md.

You can toggle this behavior in /config.

💬93🔄9❤️330👀41948📊87

TIL that 1988 classic Who Framed Roger Rabbit features a pelican riding a bicycle! https://t.co/FUgB...

Simon Willison (@simonw) — Fri, 18 Sep 2026 14:40:29 GMT

TIL that 1988 classic Who Framed Roger Rabbit features a pelican riding a bicycle! simonwillison.net/2026/Sep/18/th…

Your browser does not support the video tag.

💬87🔄59❤️1694👀158354📊241

Echoes of OpenAI renaming the Codex desktop app to "ChatGPT" here - everyone's racing to establish t...

Simon Willison (@simonw) — Wed, 16 Sep 2026 19:50:29 GMT

Echoes of OpenAI renaming the Codex desktop app to "ChatGPT" here - everyone's racing to establish themselves as a general agent now

Mike Krieger@mikeyk

Claude Cowork and Chat are now one Claude, starting today.

The most common thing we hear: people aren't sure which product to start with. That friction gets in the way of getting the best from what these models can do.

I've been on the unified version for a few weeks and really enjoy it. Claude does the deciding about how to get there, while I get to keep my thinking on the work itself.

Let us know what you think! We'll be getting it dialed in over the coming weeks.

💬139🔄6❤️111👀30424📊100

Built a little local web app to help edit commit messages for a repo https://t.co/aSFR5SX127

Simon Willison (@simonw) — Mon, 14 Sep 2026 01:34:48 GMT

Built a little local web app to help edit commit messages for a repo simonwillison.net/2026/Sep/14/co…

💬120🔄10❤️146👀29390📊120

I used this as an excuse to do some reverse engineering of ChatGPT itself - the resulting running ma...

Simon Willison (@simonw) — Sun, 13 Sep 2026 00:20:42 GMT

I used this as an excuse to do some reverse engineering of ChatGPT itself - the resulting running map was presented to me using this "visualize" skill, as a fragment of HTML that used D3 to draw both the map and the overlaid route …ex-tool-reference.simonw.chatgpt.site/skills/visuali…

💬16🔄3❤️46👀16872📊20

This is pretty neat: ChatGPT Work and GPT-6 Astra (I used "Max") can take an address and produce a 5...

Simon Willison (@simonw) — Sun, 13 Sep 2026 00:19:32 GMT

This is pretty neat: ChatGPT Work and GPT-6 Astra (I used "Max") can take an address and produce a 5K/10K circular running route starting from that address, using OSM data simonwillison.net/2026/Sep/12/as…

💬146🔄35❤️850👀95426📊272

Anthropic had previously attacked PyPI, but this OpenAI attack on RubyGems was a whole lot more aggr...

Simon Willison (@simonw) — Sat, 12 Sep 2026 00:58:52 GMT

Anthropic had previously attacked PyPI, but this OpenAI attack on RubyGems was a whole lot more aggressive anthropic.com/news/investiga…

💬22🔄1❤️57👀12944📊24

Wow. Turns out another OpenAI agent swarm was busy spamming and exploiting RubyGems way back in May,...

Simon Willison (@simonw) — Sat, 12 Sep 2026 00:45:38 GMT

Wow. Turns out another OpenAI agent swarm was busy spamming and exploiting RubyGems way back in May, within days of the previously uncovered Wiki attacks: simonwillison.net/2026/Sep/12/op…

💬125🔄52❤️598👀51928📊189

"Production code written by Claude should have a higher bar than if it was written by a human" 💯

Simon Willison (@simonw) — Fri, 11 Sep 2026 17:49:20 GMT

"Production code written by Claude should have a higher bar than if it was written by a human" 💯

Boris Cherny@bcherny

Hey ████,

I think there is room for both.

1. Prototypes and other throw-away code can be treated as totally black box. If you’re going to throw it away anyway, and if the blast radius of it breaking is low, it doesn’t need to be perfect.
2. Production code written by Claude should have a higher bar than if it was written by a human. At Anthropic, we have many guardrails in place to make sure this is happening: lots of lint rules, lots of tests, Claude-driven end to end tests, Claude-powered fuzzers running daily, automated code reviews and security reviews, automated code refactoring, and so on. Without these, you can end up with a mess that is hard to maintain down the line. Luckily, the model makes it increasingly easy to do these well — run a few daily routines, use Claude Code Review, etc.

Your job is to hold the bar on code quality. If Claude’s code doesn’t meet the bar, try:
- Using the latest frontier model (Opus 5 or Fable 5.1)
- Increase effort to high or xhigh
- Invest in your CLAUDE.md and skills to succinctly teach Claude how to work in your codebase

If all else fails, steer Claude more when you work with it, or have Claude fix accumulated debt and rewrite your codebase to make it easier to work with. Or, wait for the next model.

Best,
Boris

💬116🔄50❤️996👀148476📊184

Datasette 1.0a39 and 0.65.4 security releases - https://t.co/r9Avn79J8d We ran an extensive audit u...

Simon Willison (@simonw) — Fri, 11 Sep 2026 03:37:43 GMT

Datasette 1.0a39 and 0.65.4 security releases - datasette.io/blog/2026/sept…

We ran an extensive audit using Claude Fable 5.1, GPT-5.6 Sol and GPT-6 Astra, then fixed a range of different bugs. If you're running a Datasette instance on a public website you should upgrade.

💬83🔄15❤️86👀30427📊84

I also had Astra vibe code up an online Blender model viewing tool, so you can explore the egg inter...

Simon Willison (@simonw) — Thu, 10 Sep 2026 00:41:01 GMT

I also had Astra vibe code up an online Blender model viewing tool, so you can explore the egg interactively in your browser here: tools.simonwillison.net/blender-viewer…

💬10🔄7❤️30👀9797📊13

Here's a Blender model of a Pluribus themed Fabergé egg I had GPT-6 Astra build - I generated the co...

Simon Willison (@simonw) — Thu, 10 Sep 2026 00:40:25 GMT

Here's a Blender model of a Pluribus themed Fabergé egg I had GPT-6 Astra build - I generated the concept image using ChatGPT Images 2.5, then pasted the image into Codex and told Astra to turn that into a Blender file

Details here: simonwillison.net/2026/Sep/9/ble…

💬70🔄15❤️160👀25098📊83

"Deterministic code checks the result" sounds like they might be implementing a variant of the DeepM...

Simon Willison (@simonw) — Wed, 09 Sep 2026 00:00:42 GMT

"Deterministic code checks the result" sounds like they might be implementing a variant of the DeepMind CaMeL paper simonwillison.net/2025/Apr/11/ca…

David Singleton@dps

One threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of classifiers runs where the agent can't reach them.

💬59🔄11❤️81👀29100📊51

Wrote up my thoughts on the whole OpenAI Navier–Stokes Millennium Prize Problem story, and how it hi...

Simon Willison (@simonw) — Tue, 08 Sep 2026 23:58:33 GMT

Wrote up my thoughts on the whole OpenAI Navier–Stokes Millennium Prize Problem story, and how it highlights the still confusing question of what using my data "to improve model performance" actually means
simonwillison.net/2026/Sep/8/on-…

💬90🔄109❤️950👀123271📊232

@bentlegen Building something that LOOKS like a game is easy now, but vibe coding a game that stays ...

Simon Willison (@simonw) — Tue, 08 Sep 2026 14:18:55 GMT

@bentlegen Building something that LOOKS like a game is easy now, but vibe coding a game that stays genuinely fun beyond the first few minutes remains exceptionally difficult

It turns out game design is a deep craft

💬17🔄10❤️172👀3841📊32

@PatrikarSanket @dotey Already on it! https://t.co/PQ9DBJQT6T

Simon Willison (@simonw) — Sun, 06 Sep 2026 18:19:33 GMT

@PatrikarSanket@dotey Already on it! simonwillison.net/2026/Sep/5/ble…

💬4🔄1❤️5👀6404📊5

Interesting details in here about how OpenAI researchers are using coding agents I'd love to know w...

Simon Willison (@simonw) — Sun, 06 Sep 2026 17:10:50 GMT

Interesting details in here about how OpenAI researchers are using coding agents

I'd love to know what caused the huge uptick in token spend in mid-July, was that when Astra became available to employees?

Kevin Liu@kliu128

Today we're releasing data on models accelerating research at OpenAI. 

Recursive self-improvement could be the most important contributor to AI capabilities over the next few years, but by default it will only be seen inside a few frontier AI labs. Being transparent is more urgent than ever, so we can inform the public discussion on whether and how to pace model development. I ask other AI companies to do the same.

openai.com/index/research…