Now Playing
GPT-5.6 Cost Collapse, Anthropic's $10.9B & Pentagon Deorbit Contracts
Audio is available on Spreaker — see link below.
OpenAI's GPT-5.6 slashes agentic costs by 96%, Alibaba's Qwen beats Claude Opus on a single GPU, and Anthropic posts $10.9B in Q2 revenue two years ahead of schedule. Plus: a 181K-record AI data breach, neural net cryptanalysis at CRYPTO 2026, and the Pentagon's $8.4M commercial deorbit bet.
Audio is available on Spreaker — see link below.
The Pentagon just handed commercial space companies eight point four million dollars to solve one of the most urgent infrastructure problems in orbit. Firefly, D-Orbit, and Katalyst each received contracts to design spacecraft that can autonomously deorbit unprepared satellites.
Shifting to AI infrastructure, OpenAI shipped the GPT-5.6 family. Three variants: Sol, Terra, and Luna.
Meanwhile, Alibaba's Qwen three point eight, twenty-seven billion parameter dense model, has outperformed Claude Opus four point six Max on SWE-Bench Pro. The part worth watching is that it runs on a single twenty-four gigabyte GPU and ships under Apache two point zero.
On the security side, AI notetaking platform tl;dv left one hundred eighty-one thousand eight hundred and seventy-four meeting records queryable by any signed-up user for six months. Two million users affected.
At CRYPTO 2026, a new dedicated track formally treats trained neural networks as mathematical objects subject to cryptanalytic attack. One hundred eighty-nine papers across ten volumes, a record.
Two financial signals to close on. Anthropic posted ten point nine billion dollars in Q2 revenue, up one hundred thirty percent year over year, with five hundred fifty-nine million dollars in operating profit.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.