Capex spend on AI data centres has reached 1.2% of US GDP. That has surpassed Dotcom era spend on telecom equipment (1%) but still trails railroad spend in 1880s (6%).

We have two shorthands for our GenAI projects, "80/20" and "120/40", for "80% of human quality at 20% of human effort" vs "120% quality at 40% effort". The former is known as "AI slop", and public sentiment is coalescing around the idea that that's all GenAI is capable of. 1/

I think the 99% of people not knowing what's really coming in AI by 2030 include Sam Altman.

In a couple of years, no one will say they seriously thought you could get to AGI just by scaling up 2023 LLMs, but that was basically the consensus view for a certain crowd for about a year and a half.

We have been systematically testing lots of received prompting wisdom & for recent AI models: 🚫Threats, saying please, being insulting, & promising tips do not change average performance on challenging tasks ⛓️Chain-of-thought no longer helps even non-reasoner performance much

This is simply and obviously wrong in easily demonstrable ways, from the expansiveness of Roman citizenship to the incorporation of Persian elites under Alexander and the Seleucids (however poorly) to obvious things like the numbers of Scots in key posts in the British Empire.

1/5 I agree with Setser. What some people are seeing as "export strength" is really, in this case, demand weakness. Normally if a country's manufacturing production is surging because of increased efficiency, the "reward" for that rising efficiency should presumably come...

So many people here hate modernist architecture but I'm convinced they don't even know what it is. Oscar Niemeyer's 1964 Strick House in Santa Monica is up for sale ($16M) and it's incredible. 1911 La Mesa Dr. Santa Monica, CA 90402

DeepSeek, BYD, Unitree, Xiaomi — a few Chinese companies have been grabbing global attention. But these companies are just the tip of the iceberg. There are thousands of Chinese firms in the supply chain and in other industries that have advanced swiftly over the past decade.

What a clickbait take by ian bremmer. The FT article these figures were taken from explains perfectly well why people believe crime has risen. Spoiler alert: highly visible criminal offences like snatch-theft have risen sharply in resent years. John Burn-Murdoch

Why did Intel fail, a speculative 🧵

Invasion of the killer tomatoes. 🤣

The starkest evidence that Nvidia has become a meme stock: the fact that it went up today despite the underwhelming GPT-5 presentation, which made four things clear: • AGI is not close. • We have reached a point of diminishing returns. • No company has a clear technical

Is Chain-of-Thought Reasoning of LLMs a Mirage? ... Our results reveal that CoT reasoning is a brittle mirage that vanishes when it is pushed beyond training distributions. This work offers a deeper understanding of why and when CoT reasoning fails, emphasizing the ongoing

Letter of the day (via alan rusbridger)

It's been over six months since DeepSeek launched R1, and the world is still waiting for its successor model from China's AI darling. In the meantime, many developers have shifted to Alibaba's Qwen3 series, which adapted many of R1's core ideas but made it easier and cheaper to

‘Absolutely immense’: the companies on the hook for the $3tn AI building boom - on.ft.com/3HxLJJK via Financial Times

One could call this a huge transfer of wealth from Venture Capital LPs to NVIDIA.