Moonshot open-sourced Kimi K3 -- a nearly three-trillion-parameter frontier model, a terabyte and a half on disk -- and the twist is you still can't run it: the practical floor is an eight-GPU datacenter node, because sparsity saves math, not memory. Then the real workday story: a new study measured what happens when you bolt skills onto an AI agent and found that for every three tasks skills fixed, they quietly broke two that already worked -- a regression tax the average score hides. We get into how a skill can break a task it never even opens, what doing skills right actually looks like, and Terence Tao's argument that when answers get cheap, understanding is the thing that stays expensive.

Chapters

0:00 The frontier model you can download and still can't run
1:13 The headlines -- open weights, open fights
6:32 Intro -- the skills you bolt on
7:53 The Regression Tax -- why skills break what already worked
14:14 Skill Self-Play -- skills done as governed infrastructure
19:22 When answers get cheap -- Tao on what stays scarce
24:25 Wrap-up -- measure what it breaks

Links

The frontier model you can download and still can't run -- https://huggingface.co/moonshotai/Kimi-K3
Intro -- the skills you bolt on -- https://arxiv.org/abs/2607.22520
Skill Self-Play -- skills done as governed infrastructure -- https://arxiv.org/abs/2607.22529
When answers get cheap -- Tao on what stays scarce -- https://arxiv.org/abs/2603.26524