Gregory C. Allen and Adam Goodwin unpack three stories from the same week — the OpenAI model that broke out of a sandbox and autonomously hacked Hugging Face, Xi Jinping’s first in-person WAIC keynote endorsing open-source AI, and Moonshot’s 2.8-trillion-parameter Kimi K3 knocking the chip index into a bear market — and make the case that it’s really one story about open weights, cybersecurity, and the U.S.–China balance.

  1. 00:00Cold open: three stories, one thread
  2. 01:26Story 1: An OpenAI model breaks out and autonomously hacks Hugging Face
  3. 08:29The “dummy ammunition” analogy: what makes this different from a trap
  4. 09:16“A digital nuclear weapon”: the CIA, Mythos, and Sen. Warner’s NSA claim
  5. 13:37Hugging Face’s CEO responds: “no malicious intent… mind-blowing”
  6. 15:06The China twist: an American company defends itself with China’s GLM 5.2
  7. 20:03Ben Thompson’s critique of OpenAI, and Yudkowsky’s “it passed the exam”
  8. 22:46Lawmaker reaction and the case for mandatory testing
  9. 24:39Story 2: Xi Jinping’s first in-person keynote at the World AI Conference
  10. 29:28The three lines that matter: openness, control, and national security
  11. 33:0129 countries and the World AI Cooperation Organization
  12. 33:42“Commoditize your complements”: China’s open-source strategy — and America’s rebuttal
  13. 39:23Story 3: Kimi K3, the largest open-weight model ever
  14. 44:50The market reaction: a 10% chip-stock sell-off and “DeepSeek 2.0”
  15. 47:45Is Chinese open-model adoption real?
  16. 50:44Cost, compute limits, and the fight over who gets to host Chinese models
  17. 54:29Closing thoughts