Simon @spoltier@qoto.org
- Country
- 🇨🇭
- profile banner
- https://showyourstripes.info/
code / data wrangler in Switzerland.
Recovering reply guy. Posts random photos once in a while.
Joined Jul 2023
bReAkInG : Sydney Sweeney to star in gattaca remake
Simon
boosted
Where do the bad rainbows go?
Prism.
They get a light sentence.
It gives them time to reflect.
🌈⭐
Simon
boosted
Can LLMs reason by analogy like humans? We investigate this question in a new paper published in the Journal of Memory and Language (link below). This was a long-running but very rewarding project. Here are a few thoughts on our methodology and main findings. 1/9
@b0rk 😥✋ 🔴 No / 🔴 What
Simon
boosted
Made an outline over yesterday and today after having actually played with gpt-oss, considering abandoning this draft because gpt-5 just dropped. Should I polish it up and ship it anyways?
Simon
boosted
Here's new term of art I like: "asynchronous coding agents" to describe the category of software that includes OpenAI Codex and Gemini Jules - cloud-based tools where you submit a prompt and they check out your repo, iterate on solution and submit a pull request https://simonwillison.net/2025/Aug/6/asynchronous-coding-agents/
Simon
boosted
I felt like I learned a lot in the course of this conversation. About the way AI stands for fears about homogenization ("AI slop" == another version of "mass culture"), but also about a range of possibilities on the other side. Aaron makes a good case that AI will reward pluralism 41:20 ff. +
RE: https://bsky.app/profile/did:plc:x2xmijn2egk5g67u3cwkddzy/post/3lvnxbkzvcg2w
AI can be a tool for cultural understanding, and thus a way to shore up pluralism and liberalism. That's the counter-intuitive, but fascinating and...
Simon
boosted
If a battery such as this were productized, what materials would it likely be made of?
https://mastodon.social/@ieeespectrum/114971915076185383
(The article is so lightweight it communicates little, but it links a proper paper, which is so heavyweight it communicates little. I'm not even completely sure in what form the energy is stored in this battery. Trapped photons??! Could that possibly be right?)
@petealexharris @mcc it's not even that impressive, given that by "idle" he would have to mean "completely turned off". They are not drawing any power from the battery AFAICT. If I take a new smartphone, charge it for 30 minutes (on a fast charger) and leave it off for 30 days, it might still have a bit of charge.
Simon
boosted
Some more agentic coding learnings: Types are great, except when they are not. Some thoughts on why TypeScript can both help and harm, and why Go does not suffer from this. https://lucumr.pocoo.org/2025/8/4/shitty-types/
@mitsuhiko I feel like this may be a matter of how the AI system is set up in relation to the user. I've had some surprisingly good experience using rust with Google Jules - as long as you manage to make it compile the code and not give up until it's successful. Rust's compiler error messages very likely help.
Simon
boosted
Blindsight by Peter Watts is a science fiction book about the dangers of LLMs.
@fogus interesting, this was recently on the orange site https://distantprovince.by/posts/its-rude-to-show-ai-output-to-people/
@Weld I'm very suspicious of there being no relationship between the security pass rate and the syntactic pass rate.
To take Java as an example, the syntax is incorrect iff. the program doesn't compile. That should either count as perfectly safe (as the program never runs) or not safe. But the progress on syntax seems to have no effect on the security check curve!
Simon
boosted
Pretty decent pelicans from the new GLM-4.5 and GLM-4.5 Air models. Both models are MIT licensed, released by Chinese AI lab Z.ai this morning
https://simonwillison.net/2025/Jul/28/glm-45/
@xeiaso.net that this is in service of Slack kind of ruins it
Simon
boosted
- assuming that individual effects will predict collective effects on a totally different timeframe, the underlying assumption being that those collective effects we're genuinely interested in are just the additive result of many individual effects measured on a completely different timeframe
- timeframe choices in general which have no correspondence to the real question's timeframe
Simon
boosted
@mitsuhiko
The article you linked illustrates the issue well: countries defending their narrow interests and their formal sovereignty at the cost of the interests of the block and of actual power on the world stage. This is simply short-sighted, no matter what the orientation to growth is. Harmonization and simplification would already achieve sustained growth, without needing "net deregulation", as in an overall loss on worker's rights, environmental protection, etc.
The issue is that the 27 won't agree on how to achieve these, even if they agree that they are important.
Regarding growth: Europe will not become libertarian (and it should not do so in my opinion).
I agree with you that the general attitude towards progress is overly conservative. I do think it is more justified than in the past, as we have achieved a lot in many respects, and have more at stake as a result. It may well be that a more powerful Europe has the effect of slowing growth globally, and that may not be a bad thing overall (I would expect any kind of decentralization of power away from the US will have that effect in the short to medium term).
Simon
boosted
This week, you can get the ebook of The Terraformers for just $2.99 wherever ebooks are sold -- including indie bookseller platform Bookshop.org, along with the usual places. If you can't wait for Automatic Noodle, this will tide you over!
- Country
- 🇨🇭
- profile banner
- https://showyourstripes.info/
code / data wrangler in Switzerland.
Recovering reply guy. Posts random photos once in a while.
Joined Jul 2023
