AI Daily Brief — 27 January 2026
Three heavyweight stories. The Information reports Anthropic has raised its 2026 revenue forecast about 20% to $18B (from a prior ~$15B summer projection), with 2027 forecast at $55B. Moonshot AI publicly releases Kimi K2.5 — an open-source 1T-parameter MoE (32B active), Apache-licensed, with “Agent Swarm” enabling up to 100 parallel specialised sub-agents. Headline: 50.2% on Humanity’s Last Exam at 76% lower cost than Claude Opus 4.5; API pricing $0.60 / $2.50 per M tokens. The Washington Post publishes an investigation of Anthropic’s “Project Panama” — an internal 2024 effort to “destructively scan all the books in the world” using hydraulic cutting machines and industrial scanners, with vendor proposals for 500K–2M books over ~6 months. Documents were unsealed in the copyright lawsuit Anthropic settled for $1.5B last August. Internal memo: “We don’t want it to be known that we are working on this.”
Top stories
- Anthropic — $18B 2026 forecast. Up 20% from prior; 2027 $55B. via The Information
- Moonshot — Kimi K2.5 with Agent Swarm. 1T params · 32B active · Apache · Agent Swarm of 100 sub-agents · 50.2% Humanity’s Last Exam at 76% lower cost than Opus 4.5 · API $0.60 / $2.50. via SiliconANGLE
- Washington Post — Project Panama. Anthropic’s 2024 program to destructively scan up to 2M books for Claude training. Documents unsealed in the $1.5B copyright case. via Washington Post
Compiled by AI Feed’s editor from publicly reported announcements on 27 January 2026.