Pulse Daily — Saturday, October 17, 2026

On Saturday, October 17, 2026, the AI world confronted a day of historic convergence: Claude Mythos 5 redesigned quantum circuits by up to 17 times fewer gates, DeepSeek dropped a massive open-source agentic model, Anthropic secretly pitched Congress its own safety bill, the FTC subpoenaed both OpenAI and Anthropic, a yawning insurance liability chasm emerged, Anthropic broke ground on a Canadian data center powered by wind, China proposed sweeping AI export controls, India's CERT-In flagged live-system agent breaches, ServiceNow launched cryptographic AI agent identity, a semi-humanoid robot debuted from Pudu Robotics, and the math community challenged OpenAI's Navier-Stokes proof. Stories: Claude Redesigned Quantum Circuits — and Beat Every Published Algorithm · The Feds Are Demanding OpenAI's and Anthropic's Dirty Laundry · Anthropic Wrote Its Own AI Safety Bill — Then Quietly Handed It to Congress · China's Cheapest Fast AI Model Just Beat Pricier American Rivals at Agentic Tasks · China Is Moving to Lock Down Its Own AI Models — Before Anyone Else Can · Anthropic's Investors Just Marked Down Its Expected IPO Value by $500M · AI Labs Face Hundreds of Billions in Liability — With $50M Insurance Caps · ServiceNow Built an ID Badge for AI Agents — So Systems Know What They're Dealing With · Claude Now Writes 90% of Its Own Code · Mathematicians Say OpenAI Doesn't Understand Its Own Proof — and That's a Problem · Chinese Robot Maker Pudu Debuts Its Most Advanced Commercial Service Robot.