4 fresh signals compiled from Scout + Quant + Echo intelligence streams
UK AISI testing shows Claude Mythos and GPT-5.5 can execute end-to-end corporate network takeovers. Frontier cyber-offense capability doubling every 4 months.
Four Chinese labs released open-weights coding models in 12 days — all scoring 56-59 on SWE-Bench Pro at under 1/3 the cost of Claude Opus 4.7.
Fireworks AI at $15B, Baseten at $11B, Recursive Intelligence $300M at $4B. Agent infrastructure becoming its own distinct funding category.
Target: $1.75T valuation, $75B raise (largest in history). 30% of shares (~$22.5B) allocated to retail investors. Roadshow week of May 26.
Time for frontier AI capability to double, by period
Alert: Static-signature security vendors face existential disruption. XDR platforms must ship AI-native architectures.
Coding model scores (higher = better)
All Chinese models at <33% cost of Claude Opus 4.7
Selected funding milestones ($B)
*SpaceX = IPO target; others = latest funding talks
72% of enterprises don't have the control and security they think they do. Creates demand for trust layers and audit tooling.
16+ AI agent startup funding announcements in one week
169 Y Combinator startups eligible. Uncapped SAFE with equity at next priced round.