Now Playing
Unreviewed Code, Math Breakthroughs & China's Chip Gap | Ep 1
Audio is available on Spreaker — see link below.
Half of developers are shipping unreviewed AI-generated code — and the oversight gap is widening fast. Today's briefing covers Claude Code's 'dreaming' feature, OpenAI disproving a math conjecture, Google I/O's agentic science pivot, Alibaba's chip production shortfall, and a doubling of AI-driven supply chain breaches.
Audio is available on Spreaker — see link below.
Half of the developers in the room raised their hands. They were shipping AI-generated code.
The other development worth leading with is what OpenAI's general-purpose reasoning model just did in mathematics. It independently disproved an important conjecture.
Google I/O this week framed the question more directly than usual. The Gemini for Science package brings together LLM-based research systems under one architecture, and it signals a real pivot.
On the infrastructure side, a number came out of China this week that puts the scale of the problem in plain terms. Alibaba's T-Head division has produced five hundred sixty thousand Zhenwu AI chips in total.
One more development that deserves attention. Third-party involvement in security breaches has doubled, jumping from fifteen percent to thirty percent.
The thread connecting today's stories is the same in every case. Automation is accelerating, and human validation is retreating.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.