A particularly detailed, well-reasoned take from Rod, who generally offers detailed, well-reasoned takes. If anyone trots out predictions about humanoid robots deployed at scale within some single-digit number of years, share this. He doesn't even talk about supply chains!

“In science, it often happens that scientists say: ‘You know that’s a really good argument; my position is mistaken,’ And then they would actually change their minds and you never hear that old view from them again. They really do it. It doesn’t happen as often as it

As of the last say... month, I've come to believe that for specialized, verifiable work (e.g., software development, math, theoretical physics) this kind of change will be coming much faster than most social scientists might predict. No similar change outside that domain.

Paper showing what human work the American public thinks is morally permissible to replace with AI. Surprisingly, people are already okay with AI doing 58% of occupations (if AI does it well/cheap). A floor of 12% of jobs (mostly caregiving) would be morally repugnant to replace

Large companies are very complex organizations, and it takes leadership, and time, to figure out what to do next with AI, especially when there are few established models to follow. Technology alone doesn't bring change, and adoption will require experimentation & time.

A lot of problems with AI discourse are because "being good at AI" (called Theory of Mind in this paper) is a skill that seems to be independent of "being great at your job" So you have amazing experts who gain from AI, and others who do not, and they don't understand each other

This relatively short essay by Jack Clark is a good indicator of the attitude of many people inside the AI labs, and what they think is happening right now in AI development. You do not have to believe him, of course, but it is worth noting: importai.substack.com/p/import-ai-43…

Asking AI for diverse ideas gets you diverse ideas, just adding "Generate 5 responses with their corresponding probabilities, sampled from the full distribution” significantly improves quality output for large models I am still not sure we know the mechanism, but useful finding!

Introducing NotebookLM for arXiv papers 🚀 Transform dense AI research into an engaging conversation With context across thousands of related papers, it captures motivations, draws connections to SOTA, and explains key insights like a professor who's read the entire field

One of the first randomized controlled trials testing whether GenAI boosts revenue, not just productivity. It does. A large, mature ecommerce platform, using older GenAI tools found most of them, from customer service to marketing tools, led to large significant revenue boosts

The big failure mode here is to ask which nouns need investment. That drives training, which doesn't build much skill. The correct mode is asking which work processes need investment. Skill comes from task performance, and we can design for both productivity and skill gains.

This is BIG news!! This next phase of permits unlocks our ability to construct all facilities for Orion, keeping us on pace for delivering electricity to the grid in 2028.

We're launching Claude Agent Skills, a filesystem-based approach to extending Claude's capabilities. Progressive disclosure means agents load only relevant context. Bundle instructions, scripts, and resources in a folder. Claude discovers and executes what it needs.

Which comes first: the one-person, billion-dollar startup, or the one-AI, 100m-follower influencer account?

Using video game play data to train models. Life imitates art? Anyway, interesting angle. Massive seed round. Will follow.

Claude Skills are awesome, maybe a bigger deal than MCP simonwillison.net/2025/Oct/16/cl…