The Flux ↝

Weekly Roundup 2026-10-04

Normal Tech, Dented Porsches, Soviet Watches

LLMs

Things are moving so fast, it's hard to keep track of the zeitgeist. Is Jev old news? Are LLM math arguments still trendy? Is Anthropic still mad at the Pope?

Simon WIlliamson's 2026 (so far) post was helpful to provide some perspective. Reminding us that Openclaw's first Github commit was less than a year ago. That it's really only been the last 6 months that LLM code has made it to the level that it can be used reliably, let alone blindly, for production. That there are no real established processes or even agreements on approaches for conversational coding. Tokenmaxxing came and went in a few months. And Facebook went all in (then back out again) on an AI operating model. I'm not sure you could trust an opinion about the future of work for next week, let alone next year. The froth and foam is still too thick.

But, the AI hype cycle on LinkedIn seems to be slowly reversing. People are starting to say out loud that AI doesn't create value, people who use AI do. The slow shift towards treating AI as a normal technology is underway. This of course is absolute kryptonite to Anthropic's IPO, who's proposed valuation has them eating the entire software industry. This excellent article by Reuters lays out the insanity.

Speaking of normal, the smart folks at AI as a Normal Technology posted an excellent piece on how to think about AI risk forecasting. And in one sentence (after a fair amount of reasoned argument) they summed up my gut feel on every p(doom) conversation:

ultimately, lacking either an inductive or a deductive basis, all that forecasters can do is to make up a number, and those made-up numbers are all over the place.

Old Things

I'm not sure why this 4 minute film about a dented old Porsche stuck with me this week. I have a soft spot for old cars and old watches (and increasingly old cameras). While I drool over restomods like Alfaholics and Singer , I'd much rather have a sun-bleached survivor like this Porsche. Maybe it's why this watch had me intrigued for days, though I'm not sure why the owner was so shy about the brand. I would love a scratched up old Explorer and would relish the opportunity to rile up purists.