Knowledge catalogue
My hunch for now is that this was an ill-considered test which they didn't anticipate would be instantly spotted and cause (justified) uproar - here's hoping they decide that the 'test' isn't a good i
OK, here's a resolution - I managed to get it to think using these settings: 'thinking': { 'type': 'adaptive', 'display': 'summarized' }, 'output_config': { 'effort': 'max' } Without 'display': 'summa
OpenAI's new Euphony tool works almost exactly the same way as my Codex transcript viewer https://tools.simonwillison.net/codex-timeline?url=https%3A%2F%2Fgist.githubusercontent.com%2Fsimonw%2Fa9eb599
AI agents are already too human. Not in the romantic sense, not because they love or fear or dream, but in the more banal and frustrating one. The current implementations keep showing their human orig
scosman/pelicans_riding_bicycles I firmly approve of Steve Cosman's efforts to pollute the training set of pelicans riding bicycles. (To be fair, most of the examples I've published count as poisoning
This piece thinks that the reason I got a crap pelican riding a bicycle from Opus 4.7 is that it didn't think about it first, I'm trying to figure out if there's a way to force it to think that I've m
OpenAI has been using multiple informal names internally and externally for its new image generation model, including 'ChatGPT Images 2.0,' 'Image gen 2,' and 'gpt-image-2.' The post highlights incons
OpenAI released ChatGPT Images 2.0 today, their latest image generation model. On the livestream Sam Altman said that the leap from gpt-image-1 to gpt-image-2 was equivalent to jumping from GPT-3 to G
Claude Token Counter, now with model comparisons I upgraded my Claude Token Counter tool to add the ability to run the same count against different models in order to compare them. As far as I can tel
'I am envisioning a harmonious interplay of form and space, where each element finds its rightful place within the visual narrative. The pelican emerges as a central figure, its wings poised in dynami
Claude Opus 4.7 uses approximately 8% more tokens than Opus 4.6 when processing the same text-heavy PDF document, as demonstrated by a test comparing tokenization of a 15MB, 30-page PDF (60,934 tokens
Important to note: that 3x increase for images is entirely due to Opus 4.7 being able to handle higher resolutions. I tried that again with a 682x318 pixel image and it took 314 tokens with Opus 4.7 a
Simon Willison shared an experience attempting to use X's Community Notes feature to add clarification to his own tweet, experimenting with whether the self-noting functionality would work as intended
Release: llm-openrouter 0.6 llm openrouter refresh command for refreshing the list of available models without waiting for the cache to expire. I added this feature so I could try Kimi 2.6 on OpenRout
Simon Willison shared a tutorial on integrating Datasette instances with Google Sheets, demonstrating multiple methods including the importdata() function, custom named functions, and Google Apps Scri
TIL: SQL functions in Google Sheets to fetch data from Datasette I put together some notes on patterns for fetching data from a Datasette instance directly into Google Sheets - using the importdata()
Headless everything for personal AI Matt Webb thinks headless services are about to become much more common: Why? Because using personal AIs is a better experience for users than using services direct
Note to @AnthropicAI - much as I appreciate the public system prompts this would be so much more valuable to me as a Claude power user if you published the tool descriptions as well Since Anthropic pu
Simon Willison documents the differences between Anthropic's Claude Opus 4.6 and 4.7 system prompts, analyzing changes that Anthropic made public. The notes likely highlight modifications to model beh
The biggest challenge of using chat-based AI systems is that the details of what they can do are invisible - those tool descriptions are the missing manual, publishing them would be a huge benefit to
Agentic Engineering Patterns > Here's an example of a deceptively short prompt that got a lot of work done in a single shot. First, some background. I send out a free Substack newsletter around once a
Research: Claude system prompts as a git timeline Anthropic publish the system prompts for Claude chat and make that page available as Markdown. I had Claude Code turn that page into separate files fo
PyCon US 2026 will be held in Long Beach and will feature new dedicated tracks focused on AI and security topics. The announcement comes from Simon Willison, highlighting these expanded programming ar
The event is next month - talks are Friday 15th to Sunday 17th of May, with the date before that of tutorials and two days afterwards of sprints Californians are notoriously last-minute planners so I
Release: datasette 1.0a28 I was upgrading Datasette Cloud to 1.0a27 and discovered a nasty collection of accidental breakages caused by changes in that alpha. This new alpha addresses those directly:
Simon Willison discusses the perception that LLMs and coding agents are primarily useful for greenfield development (starting new projects from scratch) rather than for maintaining and modifying large
This year's PyCon US is coming up next month from May 13th to May 19th, with the core conference talks from Friday 15th to Sunday 17th and tutorial and sprint days either side. It's in Long Beach, Cal
A https://claude.ai/ feature I really like is you can tell it to 'clone x/y from GitHub' and it can then answer questions about a repo, or use snippets of code from that repo to help build new artifac
Tool: datasette.io news preview The datasette.io website has a news section built from this news.yaml file in the underlying GitHub repository. The YAML format looks like this: - date: 2026-04-15 body
This post compares the performance of Qwen 3.6-35B-A3B and Claude Opus 4.7 models on a creative task of generating SVG code for a flamingo riding a unicycle, likely demonstrating differences in their
Release: llm-anthropic 0.25 New model: claude-opus-4.7, which supports thinking_effort: xhigh. #66 New thinking_display and thinking_adaptive boolean options. thinking_display summarized output is cur
Simon Willison discusses results from a 'flamingo on a unicycle' test on his blog, likely comparing AI model performance including Qwen. The post appears to reference previously undisclosed or unconve
For anyone who has been (inadvisably) taking my pelican riding a bicycle benchmark seriously as a robust way to test models, here are pelicans from this morning's two big model releases - Qwen3.6-35B-
Release: datasette 1.0a27 Two major changes in this new Datasette alpha. I covered the first of those in detail yesterday - Datasette no longer uses Django-style CSRF form tokens, instead using modern
Release: datasette-export-database 0.3a1 This plugin was using the ds_csrftoken cookie as part of a custom signed URL, which needed upgrading now that Datasette 1.0a27 no longer sets that cookie. Tags
Release: datasette-ports 0.3 A small update for my tool for helping me figure out what all of the Datasette instances on my laptop are up to. Show working directory derived from each PID Show the full
Gemini 3.1 Flash TTS Google released Gemini 3.1 Flash TTS today, a new text-to-speech model that can be directed using prompts. It's presented via the standard Gemini API using gemini-3.1-flash-tts-pr
I ran the same prompt for a London Estuary accent, a Newcastle accent and an Exeter, Devon accent - all three audio files are now embedded in my blog post https://simonwillison.net/2026/Apr/15/gemini-
I'm certain this isn't the message they intended to present, but this comes across to me as a company saying 'we no longer trust in our own ability to keep your data secure' Open source is dead. That’
@pumfleet @calcom Did you see this piece by @dbreunig? He argues that the cost of locking down software through LLM analysis makes open source MORE valuable now: https://www.dbreunig.com/2026/04/14/cy
The real goldmine isn’t that Apple gets a cut of every App Store transaction. It’s that Apple’s platforms have the best apps, and users who are drawn to the best apps are thus drawn to the iPhone, Mac
I think we will see some people employed (though perhaps not explicitly) as meat shields: people who are accountable for ML systems under their supervision. The accountability may be purely internal,
Google's Gemini 3.1 Flash TTS (text-to-speech) model includes a notably elaborate or extensive example prompt, which Simon Willison highlighted as noteworthy. The post likely comments on the complexit
Zig 0.16.0 release notes: 'Juicy Main' Zig has really good release notes - comprehensive, detailed, and with relevant usage examples for each of the new features. Of particular note in the newly relea
Cybersecurity Looks Like Proof of Work Now The UK's AI Safety Institute recently published Our evaluation of Claude Mythos Preview’s cyber capabilities, their own independent analysis of Claude Mythos
datasette PR #2689: Replace token-based CSRF with Sec-Fetch-Site header protection Datasette has long protected against CSRF attacks using CSRF tokens, implemented using my asgi-csrf Python library. T
Research: Exploring the new `servo` crate In Servo is now available on crates.io the Servo team announced the initial release of the servo crate, which packages their browser engine as an embeddable l
The problem is that LLMs inherently lack the virtue of laziness. Work costs nothing to an LLM. LLMs do not feel a need to optimize for their own (or anyone's) future time, and will happily dump more a
The TL;DR is that Google engineering appears to have the same AI adoption footprint as John Deere, the tractor company. Most of the industry has the same internal adoption curve: 20% agentic power use
Thanks to a tip from Rahim Nathwani, here's a uv run recipe for transcribing an audio file on macOS using the 10.28 GB Gemma 4 E2B model with MLX and mlx-vlm: uv run --python 3.13 --with mlx_vlm --wit
SQLite 3.53.0 SQLite 3.52.0 was withdrawn so this is a pretty big release with a whole lot of accumulated user-facing and internal improvements. Some that stood out to me: ALTER TABLE can now add and
Tool: SQLite Query Result Formatter Demo See my notes on SQLite 3.53.0. This playground provides a UI for trying out the various rendering options for SQL result tables from the new Query Result Forma
I think it's non-obvious to many people that the OpenAI voice mode runs on a much older, much weaker model - it feels like the AI that you can talk to should be the smartest AI but it really isn't. If
Finally some good news from @simonw: The Kākāpō parrots of New Zealand are having a fantastic breeding season. Media 'Using coding agents well is taking every inch of my 25 years of experience as a so
I think it's non-obvious to many people that the OpenAI voice mode runs on a much older, much weaker model - it feels like the AI that you can talk to should be the smartest AI but it really isn't Jud
ChatGPT's voice mode, when directly queried about its knowledge cutoff date, reports April 2024, indicating it is powered by a GPT-4o era model rather than a more recent one. GPT-4o initially had ...
On April 10, 2026, Simon Willison posted a short note sharing a podcast clip from his ~1 hour 40 minute conversation with Lenny Rachitsky, highlighting a segment about kākāpō parrots. The kākāpō (*...
Release: asgi-gzip 0.3 I ran into trouble deploying a new feature using SSE to a production Datasette instance, and it turned out that instance was using datasette-gzip which uses asgi-gzip which was
Tool: GitHub Repo Size GitHub doesn't tell you the repo size in the UI, but it's available in the CORS-friendly API. Paste a repo into this tool to see the size, for example for simonw/datasette (8.1M
I care because understanding which index helps me understand things like what I need to submit content for crawling to, what search agents I should allow in robots.txt, how frequently I can expect the