Simon Willison — Week of 2026-04-04 to 2026-04-10#

Highlight of the Week#

Anthropic’s decision to delay the general release of their highly capable Claude Mythos model under “Project Glasswing” marks a significant turning point in the AI industry. The move underscores a massive shift in frontier model capabilities, as models evolve from generating text to autonomously chaining multiple minor vulnerabilities into sophisticated exploits, requiring a new level of security safeguards before release.

Simon Willison — Week of 2026-04-11 to 2026-04-17#

Highlight of the Week#

This week’s most striking revelation came from Simon’s infamous “pelican riding a bicycle” SVG generation benchmark, where a 21GB quantized local model (Qwen3.6-35B-A3B) unexpectedly outperformed Anthropic’s brand-new Claude Opus 4.7 flagship. Running locally on a MacBook Pro via LM Studio, Qwen generated a better bicycle frame and even won a secret unicycle backup test, leading Simon to conclude that his joke benchmark’s long-standing correlation with general model utility has finally broken down.

Simon Willison — Week of 2026-05-08 to 2026-05-15#

Highlight of the Week#

The standout development this week is Simon’s rapid adaptation to the latest frontier model capabilities, most notably releasing llm 0.32a2 to expose and visualize the new interleaved reasoning tokens of GPT-5 class models directly in the terminal. This perfectly pairs with his hands-on explorations of embedding LLM calls deeply into developer workflows, such as executing prompts via script shebangs and leveraging models to output rich HTML rather than just Markdown.

Simon Willison — 2026-05-24#

Highlight#

Today’s most resonant post is a highlighted quote from Armin Ronacher calling out the damaging rise of AI-generated “slop” in open-source issue trackers. It serves as a stark, practical reminder that while AI coding agents are powerful, developers must preserve raw, human-observed context in bug reports rather than relying on LLMs to rewrite and hallucinate root causes.

Posts#

[Quoting Armin Ronacher] · Source Simon amplifies Armin Ronacher’s frustration with a new, frustrating failure mode in open-source maintenance: AI-rewritten issue reports. Users are feeding observed bugs into LLMs (referred to as “clankers”), which spit out confident but highly inaccurate guesswork, fake-minimal repros, and irrelevant code analogies. The core takeaway is a plea to return to the basics of bug reporting: simply state what command you ran, what you expected, what actually happened, and provide the exact error log.

Simon Willison — 2026-05-21#

Highlight#

The major news today is the official announcement of Datasette Agent, merging Simon’s three years of work on the LLM library with Datasette to create an extensible, conversational AI assistant for querying data. It represents a huge milestone for his ecosystem, opening the door for users to naturally interrogate their databases and easily build custom tools using a new plugin architecture.

Posts#

Datasette Agent Simon officially announced Datasette Agent, a conversational AI interface that lets users ask questions of the data stored in Datasette. The post features a live demo using Gemini 3.1 Flash-Lite to successfully query a blog database to find a bird-watching record. He highlights a growing plugin ecosystem—including charts, image generation, and sandbox execution—and notes that tools like Claude Code and OpenAI Codex are proving excellent at writing these extensions. Looking ahead, Simon teased a major refactor for his LLM library, a Claude Artifacts-style plugin, and a personal AI assistant named “Claw” built using his older Dogsheep tools.

Simon Willison — 2026-04-06#

Highlight#

The most substantial update today is Simon’s look at the Google AI Edge Gallery, an official iOS app for running local Gemma 4 models directly on-device. It stands out as a major milestone for local AI, being the first time a local model vendor has shipped an official iPhone app with built-in tool-calling capabilities.

Posts#

Google AI Edge Gallery Simon highlights Google’s strangely-named but highly effective official iOS app for running Gemma 4 (and 3) models natively. The 2.54GB E2B model runs fast and includes features like vision, up to 30 seconds of audio transcription, and an impressive “skills” demo showcasing tool calling against eight different HTML widgets. Despite a minor app freeze bug and the unfortunate lack of permanent chat logs, Simon considers it a significant release as the first official iOS app from a local model vendor.

Simon Willison — 2026-04-09#

Highlight#

Today’s most substantive update is the release of asgi-gzip 0.3, which serves as a great practical reminder of the hidden risks in automated maintenance workflows. A silently failing GitHub Action caused his library to miss a crucial upstream Starlette fix for Server-Sent Events (SSE) compression, which ended up breaking a new Datasette feature in production.

Posts#

[asgi-gzip 0.3] · Source Simon released an update to asgi-gzip after a production deployment of a new Server-Sent Events (SSE) feature for Datasette ran into trouble. The root cause was datasette-gzip incorrectly compressing event/text-stream responses. The library relies on a scheduled GitHub Actions workflow to port updates from Starlette, but the action had stopped running and missed Starlette’s upstream fix for this exact issue. By running the workflow and integrating the fix, both datasette-gzip and asgi-gzip now handle SSE responses correctly.

Simon Willison — 2026-04-14#

Highlight#

Simon highlights a fascinating paradigm shift in AI security: treating vulnerability discovery as an economic “proof of work” equation where spending more tokens yields better hardening. This creates a compelling new argument for the enduring value of open-source libraries in the age of vibe-coding, as the massive cost of AI security reviews can be shared across all of a project’s users.

Posts#

[datasette PR #2689: Replace token-based CSRF with Sec-Fetch-Site header protection] · Source Simon has replaced Datasette’s cumbersome token-based CSRF protection with a new middleware relying on the Sec-Fetch-Site header, inspired by Filippo Valsorda’s research and recent changes in Go 1.25. This modern approach eliminates the need to scatter hidden CSRF token inputs throughout templates or selectively disable protection for external APIs. Interestingly, while Claude Code handled the bulk of the commits under Simon’s guidance with cross-review by GPT-5.4, Simon chose to hand-write the PR description himself as an exercise in conciseness and keeping himself honest.

Simon Willison — 2026-04-15#

Highlight#

The standout exploration today is Simon’s hands-on dive into Google’s new Gemini 3.1 Flash TTS API. It perfectly captures his rapid-prototyping ethos: encountering a surprisingly complex new prompting paradigm for an audio model and immediately using Gemini 3.1 Pro to “vibe code” a UI to stress-test regional British accents.

Posts#

Gemini 3.1 Flash TTS Google released Gemini 3.1 Flash TTS, an audio-only output model controlled via standard Gemini API prompts. Simon points out that the prompting guide is highly unusual, so he put it to the test by prompting for charismatic Newcastle and Exeter accents. To speed up his experimentation, he used Gemini 3.1 Pro to instantly vibe code a custom UI for the API.

Simon Willison — 2026-04-16#

Highlight#

The most fascinating takeaway today is a surprising win for local AI: a 21GB quantized Qwen3.6 model running on a laptop beat Anthropic’s brand-new Claude Opus 4.7 at Simon’s “pelican riding a bicycle” SVG generation benchmark. This result leads Simon to conclude that his joke benchmark’s long-standing correlation with a model’s general utility has finally broken down.

Posts#

Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7 · Source Simon put the day’s two major model releases—Alibaba’s Qwen3.6-35B-A3B and Anthropic’s Claude Opus 4.7—through his infamous “pelican riding a bicycle” SVG generation benchmark. Running locally on a MacBook Pro via LM Studio, the quantized Qwen model produced a better bicycle frame than Opus, and even won a “secret backup test” generating a flamingo riding a unicycle. Simon admits this breaks the historical correlation between his SVG benchmark and a model’s general usefulness, noting he highly doubts the 21GB local model is actually more capable than Anthropic’s proprietary flagship.