Dale Johnston opens the gate to his neat suburban garden on the western fringes of Newport in South Wales. There’s a trampoline and a slide on the lawn for his grandchildren, and chairs for him and his wife. He should be enjoying the light draining from the perfect autumnal sky – but almost his entire view is blotted out by a vast expanse of prefabricated panels.

The 23 metre-high grey hulk looming over him and his neighbours is a hyperscale datacentre. It is heralded by the area’s Labour politicians as a job creator, yet all Johnston can think about is what he has lost: “Before, the sun used to come around here and it was lovely.” The way residents have been treated is “disgusting”.

Johnston, who used to work for a chemical company in the city, worries what their lives will be like when the centre is switched on: the 36 sets of air blast coolers, needed to stop the chips inside overheating, most likely generate a constant drone, while regular testing of 60 back-up diesel generators, to prevent costly interruption to the flow of data, could produce harmful fumes. He already has to keep his bathroom window closed to stop the low hum from another facility on the rapidly expanding 46-acre datacentre campus that borders the small enclave of 27 houses.

As opposed to American political and corporate alignment?

At least China openly states that it requires AI to be aligned to the 12 Socialist values and have regulations that check stuff.

Meanwhile, on the other side, we have "Mecha Hitler" with Grok and Anthropic saying that they agree with 98-99% of what the US government does...

This is 3 months old... Did you miss the OpenAI release 2 weeks ago that dwarves this by 400x?

Edit... My bad, not as many posts here as I assumed, so I didn't check the age of the post

This is useful to me. Thanks for sharing!

I’ve turned off notifications on replies for this one because I suspect it will make a lot of people very angry, but looks like physics is starting to experience some of the same pressures that we’ve been seeing in mathematics.

Its certainly not super intelligent or generally intelligent. Its useful sure but it really is being way over done.

Reminds me of that cup of coffee that says caffeine helps me do stupid things faster or something like that…

AI helps me do more stupid things more efficiently 🤖

This demonstrates the difference between open-source and open-weight very well. An "open-source" model would have the training data and training instructions (basically the build script analogue) and you could examine and modify what goes into the final artifact.

Open-weights is just gratis. You can experiment to figure out what it does and do some retraining, but you're not given the tools for precise modification and can't just rebuild it without the ideological alignment.

In other news: water is wet

That’s not the same thing.

Claude in VS Code is an agent/coding harness. “Plan” mode helps the agent plan how to perform a task before it edits code.

Fukurama is a general-purpose chat frontend. The problem I’m solving is what happens during a long conversation: you branch into clarifications/tangents, useful information gets scattered across many messages, and the main thread becomes hard to work with.

For example, if I’m using AI to plan a trip, understand a legal concept, write something, research a topic, or just think through a complicated question, “Claude Code plan mode” is not an alternative to that workflow.

The overlap is basically just that both involve AI and the word “plan.” The interaction model and the problem being solved are different.

upd: updated for clarity

vscode + claude - select “plan” option.

i haven't seen any planning tools in chatgpt or deepseek or any other frontend i used. there are some similar tools but they aren't quite what fukurama offers.

what planning tools do you mean and what frontend do you use?

ai tools handle these “problems” already. its called planning. whats the point of this tool over using built in planning tools?

Get rusted clanker

Given the current tsunami of AI assisted vulnerability detection it's good to see that some teams are starting to see the light at the end of the tunnel.

So long, of course, that it's not the oncoming train.

China already won, so everyone joining the 'dead man walking' will get same faith..

I could see AI being the new interface people use to search the internet. At the very least, it's a strong possibility. Not that it'll make the answers any more accurate.

Yeah... on the flip side, I also wouldn't give any details if I was an anthropic employee hoping for an ambiguity-driven marketing boost.

Fortune magazine said he was unavailable for comment. To be honest, I wouldn't answer Fortune magazine either.

Thanks for adding that link. I had already read it before posting. Unfortunately it doesn't include any meaningful details of how that function was found (presumably because the anthropic guy didn't provide any yet).

Wikipedia already has a full write up. It looks like Fable simply found an example of a function that had a constant Jacobian determinant (the determinant of the matrix of partial derivatives), but the function was not injective (3 different values of x, y, and z give the function the same value). Basically, the function is not "smooth", so it can't be inverted.

See the video linked in one of my other replies in this thread - a bunch of your questions are answered there (not for this specific case, but for a bunch of similar ones).

I'm also way more interested in the details than the average news article on how AI can do maths now. Like how did they home in on this specific problem to solve? Destiny? Did they brute-force it and try 5000 other unsolved maths problems first? Was it AI assisted or does AI do this on its own? What's the prompt? How many tokens did it burn through? Has it been lucky by chance or would I be able to input the same thing and there's a 90% chance it'll solve it again?

This one? https://www.youtube.com/watch?v=ZJ8KThKAfbs

Edit: Huh, interesting. So there's a few things going on:

1. In some cases, LLMs have pulled together pre-existing knowledge that had not been connected before (sometimes across languages).
2. Brute-forcing is sometimes part-of it (with searches informed by existing research)
3. There are languages like LEAN that allow formal proofs to be written out and then tested logically. I guess this means that LLMs now have access to testing the logical validity of statements, which is helping to make a bunch of these most recent breakthroughs.

That last one is particularly interesting, I think, because it expands LLM capability from being language-focused (with interfaces for agentic work) to something that can actually understand (in a way) the validity of logical statements. I think it doesn't get it to an understanding of truths in general, because logic is always built from axioms, and it can't know the validity of the axioms (except by assumption). If I'm understanding things correctly)

Matt Parker (stand up maths on YouTube) has a video from the last solved problem about how they did it.