Arthur Clune

Claude does make me laugh. It is helping me write notes on linear predictors while being (using a viewed "from a galloping horse at midnight" approximation) a linear predictor.

In the rush to build data centers, companies are starting to use their own off grid gas generators. This results in much higher emissions – 60% to 80% – compared with using their local electricity grid and risks locking in longer-term fossil dependence. www.theclimatebrink....

Two unrelated posts that are related: @maggieastor.bsky.social on rethinking interfaces for coding agents maggieappleton.com/planning-age... and Marc Booker (fediscience.org/@marcbrooker) on using AI generated animations to understand systems brooker.co.za/blog/2026/10...

Paper from researchers at Harvard "AI Article Reviews Dominate Human Peer Reviews at Their Best: Evidence and Policy Choices" Does what it says and using Opus 4.8 gking.harvard.edu/peerless/ This seem to be the way things have to go. AI papers demand AI review due to volume

Because eMudhra is in the news, I am sharing a screenie of a footer of an email I once received from them in reply to my query about my PAN card (a government-issue tax identity and card).

The non-technical intro to this commentary on the OAI maths results is well worth reading. I can't follow the maths part! The author is a leading research mathematician and openly considering the prospect that humans will no longer do frontier maths research web.maths.unsw.edu.au/~davidharvey...

Not today Java. We’re busy

Amazing how quickly we went from "oh no LLMs are going to flood the Internet with insecure code" to "oh no LLMs have discovered the internet was already flooded with insecure code"

That frontier models have got off the 0% line on A20H sts1 is very impressive and like everything is about to be hill climbed really fast. It requires managing complex decisions now wrt risks that may occur in the future. Lots of well paid jobs look like this

Proper write up of the yandex/bot thing, directly from Cloudflare. No tl;dr for this one

erm.....

Ukrainian drone strikes on Yandex data centers in Russia did not cut “Russian bots” on social media by 50%. That fast-spreading rumor is misinformation. It’s based on a misleading, misread chart. A traceable series of high-profile accounts spread it, some with financial motivations to push junk.

To those implying that Kremlin-controlled social media bot/troll farms became silent after Ukraine hit a Yandex data center, posting this CloudFare chart ( radar.cloudflare.com/bots/as13238 ): Hate to disappoint you all ⤵️

(100-e)% by Jan

This goes surprisingly hard youtu.be/3TNpOD6bov8?...

I predict there will be all sorts of issues buried in these proofs, but they will not change the long term course. Like counting letters in strawberry or basic errors in code

uh oh "We show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations."

I was talking to somebody at a UK council recently and they said they’ve been overwhelmed by GenAI complaints. People sending in 30 page ChatGPT documents about hedge disputes and such. They can’t function.

the news

"This valuable resource, extending to approximately 1,380 sq.ft, has the potential for many uses, including; Fall-out bunker EMP resistant server farm Secure discrete storage A prepper’s paradise!"

Ok so basically what half these proofs show is that every beautiful and perfect result in mathematics is just another fucking Borwein integral under the hood and we never bothered calculating hard enough for the scales to fall from our eyes.

"Everything is turing complete" But I didn't expect this

This is a good thread up and down, and I agree 100%. The labs are solving major maths problems as a byproduct. They have no interest in maths itself. We collectively have no idea how to deal with this

An important result in applied mathematics just got posted. The weird part is the methods section: arxiv.org/abs/2610.06783

I am so sad about the Green Party today. Once I read about the Lamborghini driving Bolton councillor who voted against next zero, it was clear the party has a massive problem with entryism and no real way to manage it b/c of highly devolved structure.

@dougclow.bsky.social relevant to our earlier thread

China is both embracing AI more and regulating it (and social media in general) more heavily than the US/UK

“We find that on medium-length, well-defined accounting tasks, frontier AI models are now faster and more accurate than junior accountants, even the best one in our study.” Eighteen months ago they scored well below human accountants Good discussion here: www.mercor.com/blog/human-b...

I was at a very excellent AI in Research day today and my main takeaway is the massive, massive gap between people that are really into this stuff and know where the frontier is and how fast it is moving, and the majority who still think it's just a way to get a draft bit of text via chat

Having a system where we get our top researchers to spend their time writing grants with a 5% success rate rather than doing research seems less than ideal

Dan Luu has done the work to evaluate Ed Zitron's predictions Spoiler: they are all wrong

'It’s also noteworthy that Claude Science accomplished this in one shot, without any scientific oversight more sophisticated than “keep going.” '

This is really worth following the link. So much stuff. Ad profiling. Possible anthropic distillation

me! me! I know this! me! It can't.

New animation from Opus 5.5!

More and more I think maths is going to have to start again with ideas of what research is. "We're solving open problems so fast you can't keep up" is wild.

the world is very weird and a lot is changing but a thing we can all find shared comfort in: google’s iOS apps remain pathetic industry lagging dogshit