Anthropic eyes ~$1T valuation; Mythos ships, Managed Agents gain Dreaming/Outcomes, SpaceX Colossus deal signed

Anthropic is reportedly raising at a near-$1T valuation to fund compute for its newly launched Mythos model, currently in limited release. Alongside, it signed a SpaceX/Colossus training deal, raised Claude Code usage limits, rolled out an 'infinite' context window, and added 'Dreaming' (extended memory for self-improvement), 'Outcomes' (rubric+grader success definition) and multiagent orchestration to Claude Managed Agents on Opus 4.7 and Sonnet 4.6. Mythos has reportedly disrupted White House AI policy planning.