Simon @spoltier@qoto.org

in its first year, pastagang has created 5 collaborative creative tools, written 22 blog posts, released 3 podcast episodes, published 1 academic paper and given 1 talk. it has performed in 23 gigs in London, Sheffield, Karlsruhe, Ljubljana, Berlin, Barcelona, Linz, Landau, Boston and New York. through these, pastagang has accidentally earned over £200. it has also released 3 albums, 1 single and 119 recordings of jam sessions

This LLM security research from Anthropic also has important copyright implications:

https://www.anthropic.com/research/small-samples-poison

I experimented with using a more powerful AI (ChatGPT5 Pro) to attempt an open problem https://mathoverflow.net/questions/425509/sphere-with-bounded-curvature which is not quite in my area of expertise (specifically, differential geometry): https://chatgpt.com/share/68e85cba-7228-800e-8804-a0f41aa64e14 . The outcome was surprisingly nuanced, and illustrates a point I made in an earlier post https://mathstodon.xyz/deck/@tao/114501120421010793 that one has to measure the usefulness of a tool at multiple scales. In summary: AI assistance was very useful at small scales, somewhat unhelpful at medium scales, but somewhat helpful again at large scales; the problem remains unsolved, but I understand it much better now.

The problem asks, if a smooth immersed sphere in R^3 has both principal curvatures bounded in magnitude by 1, must it enclose at least a large a volume as the round unit sphere? This looked to me like a variational problem in which the round sphere would be the minimizer, so the problem naturally splits into a "perturbative" regime in which the immersed sphere is close to round, and the "non-perturbative" regime in which the immersed sphere is far from round. Lacking much geometric intuition on this problem, and knowing that most of my analysis-based toolkit was geared towards pertburative cases, I guessed that the perturbative regime would be the more important case, and chose to focus efforts there. (1/8)

"The Trump administration on Thursday proposed banning Chinese airlines from flying over Russia on routes to and from the United States, saying the reduced flight time this practice enables puts American carriers at a disadvantage." https://www.reuters.com/world/china/trump-proposes-barring-chinese-airlines-flying-over-russia-us-flights-2025-10-09/

What you might call a win-win kind of situation

"Error: Postal codes cannot contain spaces" Oh you believe that, do you

I paid a fuckton of money for my @frameworkcomputer laptop parts. I paid extra for the parts I could've saved money on elsewhere because I wanted to support the company and what they stand for. I even ported Debian to the RISC-V Motherboard for the product that made Framework Framework: The Framework 13.

Now Framework decided to sponsor Omarchy and Hyprland, the leaderships of which are not just "politically of a different opinion", they are actively toxic, hostile and highly discriminatory, to phrase it euphemistically.

Nirav Patel responded to this in the community forum, essentially doubling down on the decision to take a "big tent approach".

I want to reiterate: __This isn't about "political differences", because the political difference are actively harmful to democracy itself, and the physical safety of marginalised communities.

Someone I supported in good faith with both code and money turned around and gave this money to multiple people who'd rather see me dead for existing.

Needless to say, I feel quite betrayed and I'm really torn on wether or not I should even continue and proceed with the port for the second RISC-V mainboard, because I don't want to end up supporting the hate towards me, my wife or other people in the same situation as me.

I'm sure I'm not the only one who feels betrayed by this decision.

On the list of Youtube series I would absolutely watch but I don't think exist: gamer/economist walks around the world in some open world game, talking about the trading system. Though we'd find 90% of the time the "trading system" is "commodities just have their price randomly move every X hours" I know.

I've become one of those people who run multiple coding agents at once - here are my notes on embracing the parallel coding agent lifestyle https://simonwillison.net/2025/Oct/5/parallel-coding-agents/

Decided to live blog this morning's OpenAI DevDay announcements, since I'm in the audience https://simonwillison.net/2025/Oct/6/openai-devday-live-blog/

I published this piece as a free post. I would love to continue offering long-form essays like this for a broad audience. However, it is solely the support of my paid members that makes this possible.

If you think this is valuable work, consider becoming a member of Democracy Americana:

https://steady.page/en/democracyamericana/posts/c5d02496-98ac-4eac-a1dd-b8109b3e26c5

We are alarmed by reports that Germany is on the verge of a catastrophic about-face, reversing its longstanding and principled opposition to the EU’s Chat Control proposal which, if passed, could spell the end of the right to privacy in Europe. https://signal.org/blog/pdfs/germany-chat-control.pdf

One of the new skills required to get the most out of AI-assisted coding tools - Claude Code, Codex CLI, etc - is designing agentic loops: carefully selecting tools to run in a loop to achieve a specified goal. Do this well and you can solve many coding problems with brute force

Here's my expanded explanation of what it means to design an agentic loop, how to do it safely (while running in YOLO mode!) and kinds of interesting problems this approach can be used to tackle https://simonwillison.net/2025/Sep/30/designing-agentic-loops/

If you've been trying to figure out DSPy - the automatic prompt optimization system - this talk by @dbreunig is the clearest explanation I've seen yet, with a very useful real-world case study https://www.youtube.com/watch?v=I9ZtkgYZnOw

My notes here: https://simonwillison.net/2025/Oct/4/drew-on-dspy/

LLMs predict what a person would say if you want an LLM to be better than half of humanity, you remove the data that comes from the bottom 50% of humanity which, the bottom 50% doesn’t write much, so this is the default trajectory

I am and this seems to be a very rare trait. What's interesting is the relevant reasoning traces are not long or hard. "If you damage the brain the phenomenon of the soul weakens, if you damage discrete parts of the brain it weakens in specific ways, therefore the soul is made of physical parts".

RE: https://bsky.app/profile/did:plc:mpdezz4nkre7vyift2rttggl/post/3m2cjmrweec26

B is for BELIEVE.

https://www.patreon.com/posts/ghost-hunters-to-140257376

#horror #ghosts #ghoststories #urbanlegends #wellington #occult

ADHD culture: https://youtu.be/bYpCy61c8JE

Hastily made coffee video

https://xeiaso.net/videos/2025/hastily-made-coffee-video/?utm_campaign=mi_irl&utm_medium=social&utm_source=mastodon