The A.I. Daily
Simon Willison@simonw
I think October represents another inflection point in capabilities for coding agents and the models that drive them: Claude Opus 5.5 can now build video games that are actually fun
Simon Willison
30 storiesSimon Willison is a developer and blogger known for Datasette and his LLM tooling, who writes about hands-on use of frontier models. Recent coverage centres on Claude Opus 5.5: he calls October 2026 an inflection point for vibe-coded games, having built the multiplayer RTS Battle Drones across 12 iterations and 321 agent commits, and a Newsletters page for his blog via voice mode in ChatGPT's Codex tab. He also tests local Qwen 3.8 27B, seeks a sub-60GB open-weight MoE coding model, and shipped an LLM plugin for an OpenAI decision model.
Related topics
- Claude Opus 5.52
- GPT-6 Astra2
- Claude Code2
- Codex2
- Claude1
- Anthropic1
- ChatGPT1
- GPT-6 Astra High1
- AWS1
- Claude Opus 4.51
- GPT-5.11
- ChatGPT Work1
Simon Willison: October 2026 is the vibe game dev inflection point
Simon Willison argues Claude Opus 5.5 has markedly improved at 3D modeling, music generation and, crucially, game design, making October 2026 an inflection point for vibe-coded games. He built four games he genuinely enjoyed without reading the code, and finds GPT-6 Astra's sense of fun still behind Opus 5.5.
Simon Willison@simonw
Given the current vulnerability apocalypse, is there anyone left who thinks holding assets in cryptocurrency - where mistaken transactions are irreversible - is still a good idea?
Developer builds AI-coded RTS with audio-based multiplayer
Simon Willison released Battle Drones, an AI-assisted real-time strategy game built over 12 iterations and 321 agent commits. It runs on mobile and desktop and supports two-device multiplayer synced via audio.
Developer builds 500 clocks in one 6.3 MB HTML file with Opus 5.5
A developer who tests frontier models by having them build self-contained clock web pages had Claude Opus 5.5 create 500 clocks across several Claude chat sessions, resulting in a single 6.3 MB HTML file. The prompting process and page design are published on GitHub.
Simon Willison@simonw
"Anthropic told investors on Wednesday distillation accounted for 5 per cent to 7 per cent of revenue" If true, and Q2 revenue was $11.5B, then that's more than half a billion dollars spent by other labs to distill Anthropic's models!
Simon Willison@simonw
OK, "using Fourier epicycles" is a new twist [Quoting @fofrAI]: Argon has this amazing natural understanding of the world. It can do so many things from first principles. Here's a pelican riding a bike, using fourier epicycles. Video made by Argon. Time-synce…
Simon Willison@simonw
Shared some notes on Python 3.15, using it with uv, and features from Python 3.11 I'm excited to use now that 3.10 is no longer the minimum supported version for my own projects
Simon Willison@simonw
What's the best open weight Mixture-of-Experts LLM for coding that fits in less than 60GB of RAM? I think MoE might be necessary to get reasonably interactive speeds on the hardware I have access to - I want something faster than 12 tokens/second
Simon Willison builds blog feature using voice mode
Developer Simon Willison used the voice conversation mode in the Codex tab of the ChatGPT desktop app to build a new Newsletters page for his blog while cooking dinner. The feature, built with the GPT-6 Astra High model, included a new model, migration, view code, templates, and import functions.
Simon Willison@simonw
I released an LLM plugin for sending prompts to OpenAI's new Jev-clone decision model that's based on GPT-6 Luna
Simon Willison@simonw
Mistral can pelican now! [Quoting @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, includ…
Simon Willison@simonw
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
Simon Willison@simonw
Just published this post about how we’re going to need default hard budget caps on pretty much everything
Simon Willison urges default hard budget caps on APIs
Simon Willison argues pay-by-usage services and APIs should ship with default hard budget caps that cut off and return errors once a limit is hit, rather than just sending warning emails, to prevent coding agents and other automated services from running up thousands of dollars overnight. He notes AWS launched spending limits on September 16.
Simon Willison@simonw
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 acro…
Simon Willison@simonw
The most important thing about Sonnet 5.5 is that it's now the model that powers the free tier on - 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 [Quoting @claudeai]: A thread of early e…
Simon Willison@simonw
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
Simon Willison reviews 2026 in LLMs
In a WeAreDevelopers keynote, Simon Willison reviewed 2026 in LLMs, highlighting Claude Opus 4.5 and GPT-5.1 released in November 2025, which he said became reliable enough for daily use when paired with their coding agent harnesses.
Simon Willison@simonw
"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…
Simon Willison@simonw
I'm on stage for the keynote in ten minutes time! [Quoting @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…
Simon Willison@simonw
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 [Quoting @GergelyO…
Simon Willison@simonw
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 [Quoting @reach_vb]: Massive QoL update: ChatGPT Voice can now use plugins AND run on GPT-6 Astra, Sol & Luna, d…
Simon Willison@simonw
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
Simon Willison@simonw
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) [Quoting @OpenAI]: Please welcome GPT-6 Sol and GPT-6 Luna…
Simon Willison@simonw
Put together some notes on Jev and the new category of system one aka decision models
Pelican-Bicycle Benchmark: 2026 Update
Inspired by Simon Willison, the author reran an SVG-generation benchmark in 2026 using stronger models: 10 of 30 prompts were run through six models on OpenRouter, costing about $20 before stopping.
Generating running routes with GPT‑6 Astra
The author asked ChatGPT Work (GPT‑6 Astra) to generate 5K/10K loop routes from their address. The assistant used Nominatim and Overpass to fetch OSM data and produced visualizations plus GPX/GeoJSON files, but the exact code and steps were hidden in the UI and lost after thread compaction; the author argues systems must preserve pre‑compacted text and make it accessible via agent/tool calls for transparency.
Simon Willison@simonw
"Production code written by Claude should have a higher bar than if it was written by a human" 💯 [Quoting @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 thr…
Simon Willison@simonw
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
That is everything