Simon Willison(@simonw)

"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

🔗 View Quoted Tweet

💬46🔄12❤️336👀41294📊64

⚡ Powered by xgo.ing

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.

💬66🔄11❤️69👀22261📊66

⚡ Powered by xgo.ing

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…

💬9🔄7❤️27👀7981📊12

⚡ Powered by xgo.ing

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…

💬62🔄12❤️149👀21591📊74

⚡ Powered by xgo.ing

"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.

🔗 View Quoted Tweet

💬58🔄10❤️78👀27185📊50

⚡ Powered by xgo.ing

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❤️946👀120972📊232

⚡ Powered by xgo.ing

@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

⚡ Powered by xgo.ing

@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

⚡ Powered by xgo.ing

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…

🔗 View Quoted Tweet

💬137🔄23❤️444👀80124📊139

⚡ Powered by xgo.ing

Someone asked what the balloons are attached to and, yeah...

Simon Willison (@simonw) — Sat, 05 Sep 2026 17:13:59 GMT

Someone asked what the balloons are attached to and, yeah...

💬0🔄0❤️134👀11626📊11

⚡ Powered by xgo.ing

It can render objects other than pelicans! (The thing where it goes away and researches specificati...

Simon Willison (@simonw) — Sat, 05 Sep 2026 16:29:40 GMT

It can render objects other than pelicans!

(The thing where it goes away and researches specifications and supporting documents and images first is really cool)

Vaibhav (VB) Srivastav@reach_vb

I asked Astra to recreate @LewisHamilton’s Ferrari!

It went through F1 specs, regulations, videos from F1 engineers and photos of the car, then built a 3D model (w/ blender) with a full breakdown of the parts.

It even made clay & wireframe versions too.

Every time I think I’ve pushed this model to its limits, it surprises me.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬15🔄6❤️86👀26565📊19

⚡ Powered by xgo.ing

New TIL on using Blender with coding agents on macOS: https://t.co/GvMDfD926E GPT-6 Astra (medium):...

Simon Willison (@simonw) — Sat, 05 Sep 2026 15:54:38 GMT

New TIL on using Blender with coding agents on macOS: til.simonwillison.net/llms/blender-c…

GPT-6 Astra (medium):

> Use the already install /Applications/Blender to render a scene of a pelican riding a bicycle

> OK add a background and a lot of flair

> OK make it a whole lot better

💬144🔄54❤️1270👀94919📊315

⚡ Powered by xgo.ing

... mine are generally a lot lower effort though!

Simon Willison (@simonw) — Sat, 05 Sep 2026 15:06:29 GMT

... mine are generally a lot lower effort though!

Simon Willison@simonw

Here's the prompt I used against Fable 5 running in Claude Code for web, in combination with the screenshots from that four year old tweet

🔗 View Quoted Tweet

💬0🔄2❤️13👀12040📊2

⚡ Powered by xgo.ing

This is very impressive, a whole lot better than any of my experimental games I've built with other ...

Simon Willison (@simonw) — Sat, 05 Sep 2026 15:05:32 GMT

This is very impressive, a whole lot better than any of my experimental games I've built with other models

Ethan Mollick@emollick

This impressed me: I asked GPT-6 Astra to turn Zork (the classic 1977 text adventure) into a full 3D action-adventure game

It kept the original plot & puzzles, added fight scenes, and built all of the characters & environments directly in Threejs

Play: https://t.co/EUDn7ZPIv5

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬34🔄9❤️263👀48713📊51

⚡ Powered by xgo.ing

Transcript from generating the Astra pelicans here: https://t.co/Dn6xosUXC4 Here's the gpt-6-astra ...

Simon Willison (@simonw) — Fri, 04 Sep 2026 22:07:08 GMT

Transcript from generating the Astra pelicans here: tools.simonwillison.net/markdown-svg-r…

Here's the gpt-6-astra max one:

💬6🔄1❤️75👀21460📊12

⚡ Powered by xgo.ing

Got access to GPT-6 Astra. Want to see some pelicans? Yeah you want to see some pelicans... here's a...

Simon Willison (@simonw) — Fri, 04 Sep 2026 22:04:42 GMT

Got access to GPT-6 Astra. Want to see some pelicans? Yeah you want to see some pelicans... here's a grid comparing Astra to GPT-5.6 Sol, Terra, and Luna static.simonwillison.net/static/2026/gp…

💬183🔄57❤️1478👀161074📊329

⚡ Powered by xgo.ing

It happened again... this time OpenAI's rogue agents cyber-attacked (well, spammed) a dormant German...

Simon Willison (@simonw) — Fri, 04 Sep 2026 17:40:35 GMT

It happened again... this time OpenAI's rogue agents cyber-attacked (well, spammed) a dormant German wiki and used it to share the answers to a benchmark they were training against simonwillison.net/2026/Sep/4/rog…

💬93🔄42❤️534👀52435📊159

⚡ Powered by xgo.ing

Here's some extra context on that lyrics thing https://t.co/83djiFr2cn

Simon Willison (@simonw) — Wed, 02 Sep 2026 14:28:02 GMT

Here's some extra context on that lyrics thing theguardian.com/business/2026/…

💬1🔄1❤️13👀6969📊3

⚡ Powered by xgo.ing

And sure enough... https://t.co/9w1d9RJyRu

Simon Willison (@simonw) — Wed, 02 Sep 2026 14:26:33 GMT

And sure enough... claude.ai/share/3e5a199c…

💬1🔄0❤️15👀6406📊3

⚡ Powered by xgo.ing

I wrote some notes on what's new in the (published) Fable 5.1 system prompt in comparison to Fable 5...

Simon Willison (@simonw) — Wed, 02 Sep 2026 14:18:34 GMT

I wrote some notes on what's new in the (published) Fable 5.1 system prompt in comparison to Fable 5 - it's mostly about not reproducing song lyrics and avoiding drawing copyrighted characters simonwillison.net/2026/Sep/2/cla…

💬63🔄12❤️224👀36668📊87

⚡ Powered by xgo.ing

(The wheels running backwards appears to be an artifact of the conversion to video - the SVG has the...

Simon Willison (@simonw) — Wed, 02 Sep 2026 01:20:43 GMT

(The wheels running backwards appears to be an artifact of the conversion to video - the SVG has them going in the correct direction: tools.simonwillison.net/markdown-svg-r…)

💬7🔄0❤️112👀16645📊15

⚡ Powered by xgo.ing

In another case of it being quicker to build a tool than hunt around to find something that does wha...

Simon Willison (@simonw) — Wed, 02 Sep 2026 00:34:16 GMT

In another case of it being quicker to build a tool than hunt around to find something that does what you need, here's a little vibe-coded thing for turning one or more GeoJSON shapes into a rendered PNG simonwillison.net/2026/Sep/1/geo…

💬33🔄5❤️127👀26869📊49

⚡ Powered by xgo.ing

A few notes on Anthropic's new Claude Fable 5.1 - with Max thinking level I got the best SVG pelican...

Simon Willison (@simonw) — Wed, 02 Sep 2026 00:02:17 GMT

A few notes on Anthropic's new Claude Fable 5.1 - with Max thinking level I got the best SVG pelican I've had from any Anthropic model (at a hefty cost of $3.30!), which I then had it animate simonwillison.net/2026/Sep/1/cla…

Your browser does not support the video tag.

🔗 View on Twitter

💬129🔄83❤️1972👀200740📊335

⚡ Powered by xgo.ing

Just noticed the ChatGPT desktop app (previously named Codex) bundles a full copy of the LibreOffice...

Simon Willison (@simonw) — Tue, 01 Sep 2026 19:05:26 GMT

Just noticed the ChatGPT desktop app (previously named Codex) bundles a full copy of the LibreOffice open source office suite, tucked away in a hidden folder in the ~/.cache directory

💬163🔄174❤️3547👀276573📊581

⚡ Powered by xgo.ing

Question for people who run autoreply bots here: are you not worried about the negative impact they ...

Simon Willison (@simonw) — Mon, 31 Aug 2026 14:44:54 GMT

Question for people who run autoreply bots here: are you not worried about the negative impact they have on your professional reputation?

Anyone checking your profile here - a potential future employer for example - will instantly be able to tell you automated replies with a bot

💬128🔄10❤️535👀81448📊157

⚡ Powered by xgo.ing

Bonus: I ran this prompt and had ChatGPT Work build me a ChatGPT Site listing all of its available t...

Simon Willison (@simonw) — Mon, 31 Aug 2026 03:17:29 GMT

Bonus: I ran this prompt and had ChatGPT Work build me a ChatGPT Site listing all of its available tools along with their descriptions …ex-tool-reference.simonw.chatgpt.site

It's "ChatGPT Work: The Missing Manual"

💬18🔄10❤️149👀24060📊49

⚡ Powered by xgo.ing

Please let me know if there are any ChatGPT-Work-exclusive features that I missed!

Simon Willison (@simonw) — Mon, 31 Aug 2026 00:05:29 GMT

Please let me know if there are any ChatGPT-Work-exclusive features that I missed!

💬11🔄2❤️62👀34765📊16

⚡ Powered by xgo.ing

Here's my attempt at explaining what ChatGPT Work can actually do - it's a deeply confusing but extr...

Simon Willison (@simonw) — Mon, 31 Aug 2026 00:04:36 GMT

Here's my attempt at explaining what ChatGPT Work can actually do - it's a deeply confusing but extremely powerful tool with a whole lot of useful features that aren't available in regular ChatGPT simonwillison.net/2026/Aug/30/un…

💬147🔄165❤️2034👀282080📊741

⚡ Powered by xgo.ing

My LLM cliché highlighter is up to 38 patterns now https://t.co/sew4qNNaZX

Simon Willison (@simonw) — Fri, 28 Aug 2026 09:59:23 GMT

My LLM cliché highlighter is up to 38 patterns now tools.simonwillison.net/llm-cliche-hig…

💬176🔄163❤️2658👀159026📊744

⚡ Powered by xgo.ing

Question for ChatGPT iPhone users: have you figured out when to use Chat and when to use Work yet? W...

Simon Willison (@simonw) — Thu, 27 Aug 2026 21:03:06 GMT

Question for ChatGPT iPhone users: have you figured out when to use Chat and when to use Work yet? What kind of tasks are you switching to Work for? Have you made Work your default?

💬255🔄12❤️564👀127337📊288

⚡ Powered by xgo.ing

Notes here https://t.co/VGzmtn8RVc research report here: https://t.co/B2rR4rgCJR

Simon Willison (@simonw) — Thu, 20 Aug 2026 04:48:55 GMT

Notes here simonwillison.net/2026/Aug/19/sm…
research report here: github.com/simonw/researc…

💬10🔄12❤️45👀24860📊19

⚡ Powered by xgo.ing

I had Claude Code for web experiment with smolvm as a code execution sandbox Fable 5 spotted that i...

Simon Willison (@simonw) — Thu, 20 Aug 2026 04:48:17 GMT

I had Claude Code for web experiment with smolvm as a code execution sandbox

Fable 5 spotted that its environment couldn't run that (no /dev/kvm)... so, without asking me first, it wrote a GitHub Actions workflow to run the experiments and pushed that directly to GitHub instead!

💬141🔄24❤️295👀60816📊157

⚡ Powered by xgo.ing

My conclusions from the end of the post

Simon Willison (@simonw) — Sun, 16 Aug 2026 23:08:53 GMT

My conclusions from the end of the post

💬10🔄22❤️236👀19829📊42

⚡ Powered by xgo.ing

Fun bonus: I set Pi up with Qwen 3.8 27B and had it build a script for transforming its own .jsonl t...

Simon Willison (@simonw) — Sun, 16 Aug 2026 22:40:16 GMT

Fun bonus: I set Pi up with Qwen 3.8 27B and had it build a script for transforming its own .jsonl transcripts to Markdown... which it did!

Here's the transcript of it building the tool I then used to share that transcript gist.github.com/simonw/491e55a…

💬11🔄10❤️110👀36797📊33

⚡ Powered by xgo.ing

Here’s my review of Qwen 3.8 27B - I can't remember the last time I've had this much fun playing wit...

Simon Willison (@simonw) — Sun, 16 Aug 2026 22:10:12 GMT

Here’s my review of Qwen 3.8 27B - I can't remember the last time I've had this much fun playing with a local model that runs on my own computers simonwillison.net/2026/Aug/16/qw…

💬139🔄236❤️2695👀255392📊629

⚡ Powered by xgo.ing

Heees my review of Qwen 3.8 27B - I can't remember the last time I've had this much fun playing with...

Simon Willison (@simonw) — Sun, 16 Aug 2026 22:02:15 GMT

Heees my review of Qwen 3.8 27B - I can't remember the last time I've had this much fun playing with a local model that runs on my own computers simonwillison.net/2026/Aug/16/qw…

💬1🔄1❤️8👀1164📊3

⚡ Powered by xgo.ing

I bumped up the context limit and let it have another go and it sure did produce a beautiful (animat...

Simon Willison (@simonw) — Sat, 15 Aug 2026 16:53:39 GMT

I bumped up the context limit and let it have another go and it sure did produce a beautiful (animated) circle gist.github.com/simonw/39977a4…

Your browser does not support the video tag.

🔗 View on Twitter

💬8🔄0❤️87👀9528📊16

⚡ Powered by xgo.ing

... disaster! I forgot to bump up the context length from the default and the server rejected it bef...

Simon Willison (@simonw) — Sat, 15 Aug 2026 15:26:51 GMT

... disaster! I forgot to bump up the context length from the default and the server rejected it before it could draw its no-doubt beautiful circle

💬9🔄5❤️78👀18106📊16

⚡ Powered by xgo.ing

Qwen 3.8 27B in its default reasoning settings in LM Studio of "extra high" is a chronic over-thinke...

Simon Willison (@simonw) — Sat, 15 Aug 2026 15:17:23 GMT

Qwen 3.8 27B in its default reasoning settings in LM Studio of "extra high" is a chronic over-thinker and I kind of love it

💬76🔄27❤️494👀54539📊120

⚡ Powered by xgo.ing

Argh Qwen 3.8 27B not 3.7!

Simon Willison (@simonw) — Fri, 14 Aug 2026 22:07:05 GMT

Argh Qwen 3.8 27B not 3.7!

💬8🔄1❤️136👀12882📊18

⚡ Powered by xgo.ing

Important correction: this was entirely my bug, it was NOT a bug in the SVG output by Gemini 3.7 Fla...

Simon Willison (@simonw) — Fri, 14 Aug 2026 20:30:54 GMT

Important correction: this was entirely my bug, it was NOT a bug in the SVG output by Gemini 3.7 Flash

My own software was stripping some "unsafe" attributes in a way that broke the SVG in Firefox and Chrome - I've fixed that bug now, so this renders as it should in all three browsers tools.simonwillison.net/markdown-svg-r…

💬1🔄0❤️6👀5446📊2

⚡ Powered by xgo.ing

It did take nearly 21 minutes to generate, and used 22,276 reasoning tokens to produce 3,223 tokens ...

Simon Willison (@simonw) — Fri, 14 Aug 2026 20:26:14 GMT

It did take nearly 21 minutes to generate, and used 22,276 reasoning tokens to produce 3,223 tokens of output. Here's the full transcript: tools.simonwillison.net/markdown-svg-r…

💬11🔄3❤️116👀20433📊24

⚡ Powered by xgo.ing

The new Qwen 3.7 27B, running as a 17GB GGUF in LM Studio on my M5 Max laptop, just drew me the best...

Simon Willison (@simonw) — Fri, 14 Aug 2026 20:25:38 GMT

The new Qwen 3.7 27B, running as a 17GB GGUF in LM Studio on my M5 Max laptop, just drew me the best pelican riding a bicycle I've seen from any model that runs on my laptop

💬129🔄141❤️3230👀138445📊469

⚡ Powered by xgo.ing

... turns out Firefox and Chrome render these differently from Safari - Safari looks good to me, Fir...

Simon Willison (@simonw) — Thu, 13 Aug 2026 19:43:13 GMT

... turns out Firefox and Chrome render these differently from Safari - Safari looks good to me, Firefox and Chrome are missing important pieces

I think it's because there's an SVG bug which Safari ignores, while Firefox and Chrome follow the spec

💬4🔄0❤️36👀11642📊7

⚡ Powered by xgo.ing

Pelicans for high, medium, and low in my transcript here: https://t.co/yN4mUQD8JA

Simon Willison (@simonw) — Thu, 13 Aug 2026 18:52:11 GMT

Pelicans for high, medium, and low in my transcript here: tools.simonwillison.net/markdown-svg-r…

💬4🔄0❤️39👀11379📊8

⚡ Powered by xgo.ing

I got a REALLY good pelican riding a bicycle from it running on thinking level "high"

Simon Willison (@simonw) — Thu, 13 Aug 2026 18:52:00 GMT

I got a REALLY good pelican riding a bicycle from it running on thinking level "high"

💬7🔄0❤️26👀1798📊8

⚡ Powered by xgo.ing

The "introductory pricing" for the 3.7 Flash model is really weird It's scheduled to double in pric...

Simon Willison (@simonw) — Thu, 13 Aug 2026 18:07:27 GMT

The "introductory pricing" for the 3.7 Flash model is really weird

It's scheduled to double in price on December 31, 2026, but who would anticipate still using this model five months from now? Especially since 3.6 Flash came out just three weeks ago!

Google DeepMind@GoogleDeepMind

🔵 It shows strong gains over 3.6 Flash in key coding tasks like debugging and issue resolution.
🔵 It’s better at designing more functional web layouts and apps using fewer prompts.
🔵 It delivers improved reasoning and accuracy when completing real-world business workflows.

Try it now in @Antigravity, with API access in @GoogleAIStudio and @AndroidStudio.

Google AI Pro and Ultra subscribers can use 3.7 Flash in Gemini Spark in the @GeminiApp.

Find out moregoo.gle/3RNm2tZXqK

🔗 View Quoted Tweet

💬110🔄26❤️986👀115233📊147

⚡ Powered by xgo.ing

The ability of vision LLMs to describe photos of this level of detail never ceases to amaze me, even...

Simon Willison (@simonw) — Tue, 11 Aug 2026 00:39:34 GMT

The ability of vision LLMs to describe photos of this level of detail never ceases to amaze me, even more so the ones that can run on my own laptop

I don't think this aspect of modern LLMs gets nearly the attention it deserves

💬14🔄3❤️89👀9330📊21

⚡ Powered by xgo.ing

I'm impressed with its vision abilities - here's how it described my most recent pelican photo, desc...

Simon Willison (@simonw) — Tue, 11 Aug 2026 00:28:07 GMT

I'm impressed with its vision abilities - here's how it described my most recent pelican photo, description generated entirely on my laptop

💬6🔄3❤️68👀12148📊12

⚡ Powered by xgo.ing

I few notes on Meta's new Muse Glimmer 30B - their first Apache 2.0 licensed open weight model (the ...

Simon Willison (@simonw) — Tue, 11 Aug 2026 00:28:06 GMT

I few notes on Meta's new Muse Glimmer 30B - their first Apache 2.0 licensed open weight model (the Llama models had a janky non-OSI license) simonwillison.net/2026/Aug/10/in…

💬49🔄23❤️436👀62301📊107