Skip to content
AI News HubLIVE
Public articles 501Collected articles 569Trust 88Refresh 60 min
Health HealthySource type ResearchFull-text rights Full text allowedLast ingested 2026-11-22ID simon-willisonStatus Enabled

Personal blog; posts are public and free to reference.

Latest public articles

Muru

Graham Dumpleton, creator of mod_wsgi, has built Muru, an AI-assisted Rust server for running ASGI and WSGI Python apps in both development and production. He has documented the project in eighteen detailed posts over three weeks, arguing that human judgment still decides what is good enough.

Simon Willison's WeblogIn-site articleMuru

Ravi Vakil on Alessandro Masullo's Proposed GRH Proof: 'It Is All Human, with No AI'

Ravi Vakil announced that Alessandro Masullo has a proposed proof of the Generalized Riemann Hypothesis, which Brian Conrad has been reviewing closely for months. Vakil stressed the work is entirely human with no AI involvement, and decided to release the current draft now rather than wait for further polishing.

Simon Willison's WeblogIn-site articleRavi Vakil on Alessandro Masullo's Proposed GRH Proof: 'It Is All Human, with No AI'

Levent Alpöge Claims Riemann Hypothesis Proof Discovered with Claude

Mathematician Levent Alpöge has announced on his personal website that, working with Anthropic's Claude, he discovered a proof of the Riemann Hypothesis, claiming the method extends to the Generalized Riemann Hypothesis and potentially the Extended Riemann Hypothesis, with arguments formalized without assumptions.

Simon Willison's WeblogIn-site articleLevent Alpöge Claims Riemann Hypothesis Proof Discovered with Claude

llm-gemini 0.38.1 fixes xhigh reasoning level support for Gemini 4 Hydrogen and Neon

Simon Willison releases llm-gemini 0.38.1, a dot release that adds the previously missed xhigh reasoning level for Gemini 4 Hydrogen and Gemini 4 Neon models. Comparison shows Hydrogen at xhigh uses far more output tokens and costs more, but produces quality competitive with Gemini 4 Neon.

Simon Willison's WeblogIn-site articlellm-gemini 0.38.1 fixes xhigh reasoning level support for Gemini 4 Hydrogen and Neon

Token fraud is here and it’s BAD (YouTube)

Claire Vo describes how ChatPRD's free trial was flooded by groups reselling tokens and even running a Roblox/Fortnite map studio on her agent, leading her to wind down free trials entirely.

Simon Willison's WeblogIn-site articleToken fraud is here and it’s BAD (YouTube)

PostgreSQL 19 Released

PostgreSQL 19 arrived two months late due to AI-reported vulnerability issues, but brings a key new feature: WAIT FOR LSN for read-your-writes consistency on standbys.

Simon Willison's WeblogIn-site articlePostgreSQL 19 Released

llm-gemini 0.38

Simon Willison releases llm-gemini 0.38 with support for Google's newly introduced Gemini 4 Neon, noting that the Gemini 4 pricing lineup has become even more confusing.

Simon Willison's WeblogIn-site articlellm-gemini 0.38

datasette-arrows 0.1a0

Simon Willison released datasette-arrows 0.1a0, the first plugin to use Datasette's new streaming export hook, letting tables be exported as Apache Arrow IPC streams for DuckDB, pandas, Polars and similar tools.

Simon Willison's WeblogIn-site articledatasette-arrows 0.1a0

Quoting Daniel Stenberg

curl creator Daniel Stenberg notes that while AI coding agents converted 185,000 lines of C into 340,000 lines of Rust over a little more than three weeks, the real curl project merged 300+ commits, shipped 200+ bugfixes and fixed several vulnerabilities that the rewritten clones lack.

Simon Willison's WeblogIn-site articleQuoting Daniel Stenberg

External sandboxes for coding agents

Simon Willison and Prime Radiant introduce the external sandbox pattern, where agent-executed code runs in a dedicated sandbox while the agent stays outside, releasing three server implementations supporting Pi, Claude Code, and Codex to safely test local models and defend against prompt injection.

Simon Willison's WeblogIn-site articleExternal sandboxes for coding agents

GPT‑6.2 Astra and GPT‑6.2 Sol

OpenAI releases GPT‑6.2 Astra and GPT‑6.2 Sol. Astra targets hardest projects at 80% of predecessor's price with cheaper caching; Sol keeps low pricing for everyday work. Early benchmarks show Astra trailing Claude Fable 5.2, and users report confusing communication style.

Simon Willison's WeblogIn-site articleGPT‑6.2 Astra and GPT‑6.2 Sol

Running the full server-side Datasette Python app in the browser with service workers and Pyodide

Simon Willison releases a complete rewrite of Datasette Lite using service workers and Pyodide, making the browser-based Datasette almost indistinguishable from the real server-side application, with new features like selecting SQLite versions and installing plugins.

Simon Willison's WeblogIn-site articleRunning the full server-side Datasette Python app in the browser with service workers and Pyodide

llm 0.37.3

LLM 0.37.3 is a bug fix release addressing a problem with the latest Homebrew package, where SQLite 3.54.1's corrected behavior broke LLM's logging code. The workaround wraps the subquery in an explicit json() call.

Simon Willison's WeblogIn-site articlellm 0.37.3

datasette-maplibre 0.1a0

Simon Willison released datasette-maplibre 0.1a0, a new Datasette geospatial plugin combining MapLibre GL and OpenFreeMap, built mostly from a single prompt to Claude Fable 5.2 with features like GeoJSON SVG previews and lazy-loaded WebGL.

Simon Willison's WeblogIn-site articledatasette-maplibre 0.1a0

datasette 1.0a43

Simon Willison shipped datasette 1.0a43, with the 1.0 milestone now 53% complete. The release adds streaming support for output renderers, switches built-in TSV and CSV exports to that mechanism, fixes a security issue that could expose the title and SQL of private stored queries, and introduces CSS cross-document view transitions that make the multi-page app feel like an SPA.

Simon Willison's WeblogIn-site articledatasette 1.0a43

datasette-apps 0.4a0

datasette-apps 0.4a0 adds URL fragment linking for sandboxed apps inside Datasette and reports supported URL schemes to Datasette Agent so it can help write apps that link to each other.

Simon Willison's WeblogIn-site articledatasette-apps 0.4a0

datasette-agent 0.6a0

Simon Willison releases datasette-agent 0.6a0, an LLM-powered agent plugin for Datasette that now adds an agent prompt to the homepage with hash-mention support for tables and queries.

Simon Willison's WeblogIn-site articledatasette-agent 0.6a0

datasette-llm 0.1a11

datasette-llm, the intermediary plugin for configuring LLMs in Datasette, releases 0.1a11 with support for OpenAI-compatible endpoints, enabling local models via LM Studio, Ollama, or llama.cpp.

Simon Willison's WeblogIn-site articledatasette-llm 0.1a11

llm-gemini 0.37

Simon Willison releases llm-gemini 0.37 and analyzes Google's new Gemini 4 Hydrogen model. Hydrogen is only 25% cheaper than Carbon and Argon currently, but that discount grows to 75% after December 1st. However, its output quality is noticeably lower, and it remains 15x more expensive than Claude Haiku 5.5 and GPT-6 Luna.

Simon Willison's WeblogIn-site articlellm-gemini 0.37

Gemini Decisions API

Google launches Gemini Decision 1, joining Jev and OpenAI in the decision model category. It supports yes/no, choice, and rubric-scoring questions across text, images, audio, video, and PDFs, priced like GPT-6 Luna at 10 cents per million input tokens with free output.

Simon Willison's WeblogIn-site articleGemini Decisions API

AWS ships official aws-emulator, built by an agent swarm with 29,684 commits in three weeks

AWS has released aws-emulator, an official local emulator that lets developers test against clones of many AWS services without provisioning a paid API key. The Rust-based project was developed by a large swarm of AI agents, producing 29,684 commits in its first three weeks. It currently requires Docker on macOS and Windows.

Simon Willison's WeblogIn-site articleAWS ships official aws-emulator, built by an agent swarm with 29,684 commits in three weeks

Browse tools.simonwillison.net by tag

Simon Willison added tag-based browsing to his collection of 244 vibe-coded tools. Using leftover Fable 5.2 credits, he employed tree-sitter to analyze each tool's JavaScript APIs, libraries, and web API hosts, generating tag clouds by category on the homepage.

Simon Willison's WeblogIn-site articleBrowse tools.simonwillison.net by tag

llm 0.37.2 Released: Fixes Two Security Issues That Could Leak OpenAI API Keys

Simon Willison released llm 0.37.2, fixing two security vulnerabilities reported by Daniel de Araujo that could accidentally send configured OpenAI API keys to third-party OpenAI-compatible endpoint providers.

Simon Willison's WeblogIn-site articlellm 0.37.2 Released: Fixes Two Security Issues That Could Leak OpenAI API Keys

Anthropic's IPO roadshow video

Anthropic's IPO roadshow video reveals previously unreleased details about its multi-chip training strategy and compute multipliers, with co-founder Tom Brown explaining how Claude runs on AWS Trainium, Google TPUs, NVIDIA GPUs, and AMD GPUs.

Simon Willison's WeblogIn-site articleAnthropic's IPO roadshow video

llm-gemini 0.36

Simon Willison releases llm-gemini 0.36, adding support for Google's Gemini 4 Carbon, now available to all. The model is priced at $2/$10 per million tokens until December 1st, when it doubles. Willison tests it with pelican prompts and a Three.js mech animation, plus a community demo of a 3D game built with Antigravity's /swarm.

Simon Willison's WeblogIn-site articlellm-gemini 0.36

Introducing Our Smallest, Most Efficient Model: Muse Ember

Meta AI launches Muse Ember, a small and efficient model competing with Claude Haiku and GPT Luna on price and efficiency, sitting on the Pareto frontier for cost per task.

Simon Willison's WeblogIn-site articleIntroducing Our Smallest, Most Efficient Model: Muse Ember

Quoting Prompting Claude Fable 5.2

Anthropic documentation reveals that telling Claude Fable 5.2 it owns the codebase significantly boosts its tendency to merge duplicate code, from under 10% to about 50% in tests.

Simon Willison's WeblogIn-site articleQuoting Prompting Claude Fable 5.2

Yonkers Rises

Simon Willison used Claude Fable 5.2 to build a Three.js 3D animation in which the Yonkers city boundary stands up as a giant mech and stomps across Manhattan, Brooklyn, Queens, and the Bronx.

Simon Willison's WeblogIn-site articleYonkers Rises

llm-anthropic 0.31 Released with Support for Fable 5.2 and Mythos 5.2

Simon Willison released llm-anthropic 0.31, adding Anthropic's newly launched Fable 5.2 and Mythos 5.2 models to his LLM command-line tool. Testing the classic pelican-on-a-bicycle SVG at all five reasoning levels showed xhigh producing the best result, while max burned 128,000 tokens thinking and returned no image—a bug previously seen with Opus 5.5 but new to the Fable series.

Simon Willison's WeblogIn-site articlellm-anthropic 0.31 Released with Support for Fable 5.2 and Mythos 5.2

Locating the source of a mysterious beep with Claude

Simon Willison used Claude Code to build Beep Locator, a system that turns spare phones, tablets, and laptops into distributed microphones to triangulate a mysterious beep, pinpointing it within about 40cm and finding a carbon monoxide detector in need of new batteries.

Simon Willison's WeblogIn-site articleLocating the source of a mysterious beep with Claude

All sources