Ep. 56 — Hugging Face puts the open-model commons up for auction, Thomson Reuters swaps some of Claude for a realigned Chinese model, and OpenAI's agents hit a one-percent wall

Hugging Face puts the open-model commons up for auction, Thomson Reuters swaps some of Claude for a realigned Chinese model, and OpenAI's agents hit a one-percent wall 00:48 — Hugging Face is exploring a sale that could value it at $13 billion or more - the open-model commons goes up for auction 02:17 — Thomson Reuters ships Thomson-1, built from a realigned Alibaba Qwen model, to lean less on Claude - and frames it as buying instead of renting 03:47 — OpenAI is building an agent for everything - and the adoption split is 20 million versus a billion, with under 1% of individual subscribers touching Codex 05:17 — A study says frontier labs still won't say how they'd contain a rogue model - the same weekend OpenAI asked California to make its safety law stricter 05:42 — An NBER paper finds workers are retroactively adding AI skills to jobs they already left - and a 2026 snapshot overstates 2022 AI skills by about 30% 06:06 — A watermark remover built in five hours went viral - and its author says the problem is false positives, not attribution 06:30 — Kids learn language on about 100 million words; frontier models take five orders of magnitude more, and nobody can explain the gap Daily Prompt with Archer & Iris.