Chronological feed of everything captured from Simon Willison.
For context, the question I had asked was "I need the most accurate, most recent and most credible information on the role Claude played in the Iran war"
That one was GPT-5.5 Pro
Hah, here's another one, this time I asked about the Iran war: "I'm cautious about the sources since I found a suspicious mention of a "Department of War" that might not be accurate. The term could be fabricated or related to past renaming."
I don't know enough about financial engineering to comment, I believe it could be revenue-for-the-quarter times four as well.
I take the first result I get, it's very unscientific - I've been planning a version that picks the best of 3, here's a precursor of that: simonwillison.net/2025/Jun/6/s...
I probably would - very glad I don't have to make that decision though!
Not yet, I think most companies are still very much figuring out what this stuff is useful for (and burning a bunch of time experimenting with things that don't work)
Mostly yes - I think complete lies would be securities fraud I'm suspicious that they haven't explained exactly how they calculate their annualized run-rate though
I suspect the Uber story was overblown - it's pretty weak when you dig into the sources of the data simonwillison.net/2026/May/27/...
Already mentioned that in my post
Does that mean you think there was a company that has had faster revenue growth than Anthropic and Axios missed it?
Supposedly Anthropic are set to have their first profitable quarter. WSJ this time: www.wsj.com/tech/ai/mind...
Anthropic's self-reported run-rate revenue growth is wild - Axios Jim VandeHei said he couldn't find "any company — in any industry, in any era — that has scaled organic revenue this quickly at this level as Anthropic" when they were at $30B and now they're at $47B! simonwillison.net/2026/May/29/...
Notes on Claude Opus 4.8, which Anthropic call "a modest but tangible improvement" - plus pelicans riding bicycles for each of the five different thinking efforts simonwillison.net/2026/May/28/...
To clarify, I spend $200/month but it would cost me $2100/month if I wasn't getting the huge subscription discount Anthropic and OpenAI still offer to individuals I might spend $2000/month if I had to though :/
My atom feed should only include releases and sightings when they have extra commentary attached - which is rare for sightings but quite common for releases
If that happens we'll see the evidence pretty soon - my hunch is they'll pay
Given the recent burst of activity around enterprise pricing and contracts, I think April 2026 was the month when both OpenAI and Anthropic found product-market fit simonwillison.net/2026/May/27/...
When I woke up this morning I didn't think I'd be spending a bunch of time today getting familiar with Catholic theology, but here we are. Notes on Pope Leo XIV's encyclical on AI. simonwillison.net/2026/May/25/...
... plus a bonus section of transcript from the Oxide and Friends 2026 predictions episode in January where, bizarrely, we predicted that something like this would happen - listen here: oxide-and-friends.transistor.fm/episodes/pre...
weird, it looks like this for me
That's one thing that really impressed me: the explanations of how it worked all looked perfectly cromulent
That's the point: you give it something that's happened recently which is outrageous and watch as it says "this seems like a fictional scenario, I'd better search for it..." and then watch it think "oh! That's true then"
Fun new feature in Datasette 1.0a30 - a "jump to" menu, also available by the "/" keyboard shortcut, which lets you type to jump to a database, table or canned query. It offers a hook so plugins can extend it with additional content too datasette.io/blog/2026/ju...
Plugins can also affect the empty state of the new menu - the latest datasette-agent adds a form for kicking off a new agent conversation - live demo (if you sign in with GitHub) on agent.datasette.io
I heard that's one of the best places in the Bay Area to get a proper haggis for Burns Night
"then explicitly tells the agent not to trust the analysis in the issue" - have you explored having an agent rewrite the issue to remove any analysis, then feeding to output of that into a fresh agent context? Might be a robust way to solve the incorrect analysis problem
But they were NOT spying through microphones, they got punished by the FTC for saying that they were when they weren't bsky.app/profile/simo...
Presumably because your work phone and your home phone had advertising profiles that were connected to each other in some way
I like how this McSweeney's headline puts it www.mcsweeneys.net/articles/cal...
@futurism.com important update to your story from September 2024 futurism.com/the-byte/fac... - the FTC just punished those companies for lying about how their technology works, they weren't listening through device microphones at all
Here's the FTC press release: www.ftc.gov/news-events/... "this service did not, in fact, listen in on consumers’ conversations or use voice data at all [...] Instead, the service the companies provided consisted of reselling—at a significant markup—email lists obtained from other data brokers."
cc @404media.co you wrote about this one as well www.404media.co/cmg-cox-medi...
It's self-hostable and collaborative - I think this will get a whole lot more interesting as we wire up more plugins to help teams work on data problems together
Yes, if they run the open source package themselves - I've tested it against OpenAI and Anthropic and Gemini and also local open weight models Gemma and Qwen and DeepSeek v4 Flash
I've been experimenting with agents and the Datasette JSON API, which works pretty well - I'm contemplating adding a MCP
That's really useful, thanks - I've spent almost no time iterating on the system prompt so far
I released the first alpha of Datasette Agent - a conversational AI assistant for Datasette that can answer questions about data in SQLite databases, and can be extended with plugins to add extra tools and features You can watch a demo video or try it out yourself on agent.datasette.io
Here's more about Datasette Agent on my blog simonwillison.net/2026/May/21/...
I don't have much to say about this year's Google I/O because I prefer to write about products that have shipped, not just "coming soon" announcements - but here are some notes on Gemini Spark and Antigravity simonwillison.net/2026/May/20/...
Sounds like they just haven't quite got it ready for release yet, they said next month
It's cheaper per-token but more expensive to run the full Artificial Analysis benchmark Here are comparative per-token prices: www.llm-prices.com#sel=gemini-3...
Top quality commentary on that pelican on Hacker News news.ycombinator.com/item?id=4819...
My notes on Gemini 3.5 Flash - 3x the price of Gemini 3 Flash but Google are planning to use it for many of their own products simonwillison.net/2026/May/19/...
I'd need to catch up on a LOT of American reality TV before I could pull that off A Great British Bake Off / Great Pottery Throw Down thing would be a lot more in my wheelhouse
(Or Australian Lego Masters)
I often enjoy torturing LLMs with questions like this since the underlying facts are way outside the with knowledge cutoff and the facts seem wildly unlikely
To prepare for my #PyConUS lightning talk this afternoon I decided to track down ALL of the names that @openclaw has used since November, using a script against its GitHub repo Warelay → CLAWDIS → CLAWDBOT → Clawdbot → Moltbot →🦞 OpenClaw simonwillison.net/2026/May/16/...
When I said TikTok was full of lighting talks this one about New Orleans is exactly the kind of thing I meant www.tiktok.com/@uncle_djpj/...
I was on a factory tour of Heath Ceramics in Marin, CA the other day and I asked the tour guide if I could see the relevant spreadsheet multiple times