Now Playing
Coinbase's Chinese AI Bet, Kimi Beats GPT-5.5 & the Enforcement Gap
Audio is available on Spreaker — see link below.
Coinbase slashed its AI costs 50% using sanctioned Chinese models — and there's no law stopping them. Today's briefing covers the Kimi K2.6 benchmark upset, Beijing's access restrictions, Claude Opus 4.7, and the frontier price war nobody is winning.
Audio is available on Spreaker — see link below.
Coinbase just cut its AI spending in half. The models it used to do that are on the US Commerce Department's restricted list.
The cost story would be easier to contain if these models were just cheap. They're not just cheap anymore.
That's where the Booz Allen study lands with real force. The report frames Chinese coding model adoption the way the US framed Huawei in telecoms.
While Washington struggles to close the enforcement gap, Beijing is moving in the opposite direction. The Ministry of Commerce held meetings with Alibaba, ByteDance, and Z.ai to discuss restricting overseas access to advanced Chinese models.
The commercial pressure isn't easing anywhere. Grok four point five, GPT five point six, and Meta's first paid model all launched in the same week.
Two other developments worth tracking. Anthropic released Claude Opus four point seven with improved coding and visual intelligence, priced identically to its predecessor.
The signal here isn't just one company saving money on AI. It's that the regulatory framework around Chinese model adoption is being stress-tested in public, by a major US financial institution, with a named cost figure attached.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.