This is even more dumb than it sounds. The "super smart" Balaji is letting "and so on and so on" do most of the work in his 'arguments'. It's also staggeringly boring. The future of media is here.

These days I think the biggest tell is when every “firefighter” looks like an A list Hollywood star post glow-up

This is the mathematician equivalent of screaming and running around

I hope you got danger money for that!

A screenshot of a white-on-black terminal depicting a 19x19 go board in ascii graphics, with empty grid intersections as periods, and black and white as Os and #s

It’s absolutely incredible that one of the largest Japanese-run Go servers, which has been running since 1992, is still accessed entirely via Telnet. And while most players use GUI clients that use Telnet under the hood, you can still connect manually and get ASCII graphics streamed to you

Hooks reference - Claude Code Docs Reference for Claude Code hook events, configuration schema, JSON input/output formats, exit codes, async hooks, HTTP hooks, prompt hooks, and MCP tool hooks.

The only reliable way of stopping this is with a hook on the pre-tool-use that blocks code.claude.com/docs/en/hooks and www.the-agentic-engineer.com/blog/2025-10...

Claude 4.7 has just run non-stop for a little under 45 mins making a bunch of changes to my code (using @s.ly Superpower plugin). They are good changes as well.

Shit is getting scary.

The AI Revolution in Math Has Arrived | Quanta Magazine AI is being used to prove new results at a rapid pace. Mathematicians think this is just the beginning.

Can confirm the headline is backed by the text

My bad. Make that 'white' and it's 6,400. Which is way smaller than I would have expected across y&h, but I'm sure that's my expectations being wrong.

I'm not convinced there's only <100 of me

Hederson linking AI deskilling in education to a study on colonosopists

Talking to agent is wierd. This actually makes sense in the context of what it's been finding this week

Before Christmas I wasn't using much. Now the tokens I'm using on an enterprise plan would be 10x if I paid API costs. Which I can see is going to be an issue for Anthropic

Has it been obvious for a long time? All the major companies have said they were net positive on models excluding the large cost of training the next one.

My guess is that it's changed since Jan with the models getting good, code working and agents taking off.

AI Use Appears to Have a "Boiling Frog" Effect on Human Cognition, New Study Warns A new study claims to offer the first causal link between AI dependency and cognitive erosion. Researchers warn of long-term implications.

NEW: A new pre-print study offers the first causal evidence that outsourcing reasoning tasks to AI can rapidly erode users' independent performance AND their will to persist despite difficulty.

"People’s persistence drops... they’re also not willing to try without AI.”

futurism.com/artificial-i...

I'm tweaking a presentation. So far this involves Claude Code re-writing my custom .md to reveal.js pipeline and it installing random stuff via homebrew.

It's fine.
It's all fine.

China shock 2.0: the flood of high-tech goods that will change the world Impelled by furious competition, hefty subsidies and sheer scale, the country’s companies are cutting a swath through the world’s most advanced industries

Great piece.

These are wild examples of rejecting students

(original article is from The Times)

Makes sense. Maybe for pentesting it’ll come down to how much will be “let best model guide the whole thing” v “use local models to write scripts/small codes”?

I don’t really follow the logic here. I may be missing something. Are you expecting local models to get both good enough and small enough? My expectation is that we’ll continue to use more smarts to get more done, and that this continues for a while - open models 6-12 months behind but also big

Friday afternoon posting - after a gap, I've written another blog post. This time on spec driven development and the confusion about what it is

Friday afternoon posting - after a gap, I've written another blog post. This time on spec driven development and the confusion about what it is

itym "has helped destroy"

So it is. I'm so used to things being linked directly from the text that I've never noticed that

Hi Arthur. I agree it isn't very clear, but the paper is linked to at the top right of the article (see pic).

For the avoidance of doubt, it is this one: www.sciencedirect.com/science/arti...

The EU site is giving a bad gateway error now as well

Research Professional seems to have a terrible policy of not linking to sources, but I assume it’s this paper

research-and-innovation.ec.europa.eu/knowledge-pu...

Note the cut off of 2022 means this is mainly not about LLMs

Incoming flames in 3, 2, …..

Henderson is back up.

I so admire your values. Being there for your kid is the most important thing and I'm sure that crazy sums of money were offered to not be there. Congratulations!