Surprising result from OpenAI: one of their research models achieved a gold medal performance in this year's International Mathematical Olympiad /without/ using tools
Just a classic next-token-predicting LLM with a bunch of reinforcement learning layered on top
https://simonwillison.net/2025/Jul/19/openai-gold-medal-math-olympiad/
It is tempting to view the capability of current AI technology as a singular quantity: either a given task X is within the ability of current tools, or it is not. However, there is in fact a very wide spread in capability (several orders of magnitude) depending on what resources and assistance gives the tool, and how one reports their results.
One can illustrate this with a human metaphor. I will use the recently concluded International Mathematical Olympiad (IMO) as an example. Here, the format is that each country fields a team of six human contestants (high school students), led by a team leader (often a professional mathematician). Over the course of two days, each contestant is given four and a half hours on each day to solve three difficult mathematical problems, given only pen and paper. No communication between contestants (or with the team leader) during this period is permitted, although the contestants can ask the invigilators for clarification on the wording of the problems. The team leader advocates for the students in front of the IMO jury during the grading process, but is not involved in the IMO examination directly.
The IMO is widely regarded as a highly selective measure of mathematical achievement for a high school student to be able to score well enough to receive a medal, particularly a gold medal or a perfect score; this year the threshold for the gold was 35/42, which corresponds to answering five of the six questions perfectly. Even answering one question perfectly merits an "honorable mention". (1/3)
https://halone.within.lgbt need to test something, thanks!
Boost for reach!
I scraped the schedule for Open Sauce 2025 this morning and built an alternative schedule interface with the option to add everything to your calendar (via ICS)... working entirely on my iPhone, using OpenAI Codex and Claude Artifacts
I guess you could call this "vibe scraping"? OpenAI Codex turns out to be great at writing custom scrapers if you give it internet access and tell it to download and install Playwright
Prompts + transcripts: https://simonwillison.net/2025/Jul/17/vibe-scraping/
@timkellogg.me love the golang burn
Notes on Voxtral - the new audio-and-text-input models released by Mistral yesterday. They're open weight (Apache 2) and also available via Mistral's API, so I added support for them to my llm-mistral plugin https://simonwillison.net/2025/Jul/16/voxtral/
I've been thinking about this comment from Ted a lot since he posted it. First of all, he seems entirely right that creating a system with independent goals and (the equivalent of) emotional states but with no real rights is monstrous (cont'd) /
RE: https://bsky.app/profile/did:plc:565ebob5f6hw33hjdkxty6qj/post/3ltq3xtqtjc2s
@rtyler @bert_hubert I do think a lot of people fail to grasp the potential implication of performance at scale (Dan Luu has written on this, e.g. https://danluu.com/algorithms-interviews/#appendix-misaligned-incentive-hedgehog-defense-part-3. then again, most people (me included, I expect*) do not work on code that runs at scale.
*at least that was not meant to run at scale
@timkellogg.me add overflow so True +True = False
https://www.europesays.com/fr/234842/ À Lausanne, Photo Elysée expose un travail sur le papet vaudois – rts.ch #arts #ArtsAndDesign #ArtsEtDesign #ArtsEtDivertissement #ArtsVisuels #CantonDeVaud #Culture #Design #Divertissement #Enquête #Entertainment #ExpositionArtistique #FR #France #Lausanne #PhotoElysée #photographe #Photographie #RomainMader #Suisse #Tourisme
@akamran @noplasticshower I want a t-shirt that says "a cross between a lentil and a velociraptor"
@noplasticshower "Lone star ticks are aggressive and can speedily follow a human target if they detect them. “They will hunt you, they are like a cross between a lentil and a velociraptor,” said Sharon Pitcairn Forsyth, a conservationist who lives in the Washington DC area."
@nic221 newspapers are not universal enough (Spotify/Netflix*) or specific enough (substack newsletters with a dedicated audience of 'true fans') to work as a subscription business. I personally would like to be able to pay a small amount per article, instead of all-you-can-eat or nothing, but I'm not sure if this is a widespread desire. For now newspapers are mostly losing the game of chicken with search engines (they make the whole content of the article available to crawlers, so I can read it for free on https://archive.today et al). In my home region they don't do it, which I imagine leads to very low organic growth, but maybe they don't care.
* seems to be showing its limitations with the proliferation of TV subscriptions...
code / data wrangler in Switzerland.
Recovering reply guy. Posts random photos once in a while.