Now Playing
Fable 5 Export Ban, DeepMind's Speed Leap & IPO Season Begins
Audio is available on Spreaker — see link below.
The US government pulled Anthropic's Fable 5 offline mid-launch — here's what the export control precedent means for frontier AI. Plus: DeepMind's 1,000-token-per-second breakthrough, dual IPO filings, and defense tech funding smashes its own record.
Audio is available on Spreaker — see link below.
The US government forced Anthropic's most capable model offline six days after launch, and as of June fifteenth, nobody knows when it comes back. That's the lead.
While Anthropic navigates government intervention, DeepMind shipped something that quietly reframes the competitive landscape. DiffusionGemma is a twenty-six billion parameter open-weight model that generates over a thousand tokens per second on an H100.
OpenAI confidentially filed for IPO with the SEC, joining Anthropic's earlier submission. The timing puts both companies in motion alongside the SpaceX listing.
Defense tech funding hit fourteen point six billion dollars in the first five months of twenty twenty-six. The full-year record for twenty twenty-five was nine point six billion.
Two open-weight releases this cycle reinforce the inference efficiency theme. Moonshot's Kimi K2.7-Code cuts reasoning tokens by thirty percent while improving coding performance by over twenty-one percent and multi-language work by over thirty-one percent.
Three threads carry forward from here. The Fable 5 standoff needs a resolution mechanism, and one hasn't emerged yet.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.