Phases 0–9.5 shipped the structural surgery from v3 to v4: sponsors are the witnesses, the AI Companion accompanies without authority, a single 5% voluntary donation funds the platform, the Companion is available on the daily-logging surface with voice-annotated video as the default pattern for visual practices, self-serve auth recovery and service-client migration have landed, and rooms are now the primary social surface. Phase 10A (2026-04-27) shipped Companion v2's foundations — cross-commitment memory and the Memory Curator agent — with prompt design grounded in a 3-candidate × 6-history dry-run matrix. What remains is Companion v2 Stage B (the architectural rewrite of the trigger model), Whisper voice input, optional response streaming, and a video recut of two onboarding videos.