docs: scrub rig-identifying details from the TUI sample (machine/session ids) and the provider mention
docs: #45 — one daemon per machine at both ends (atomic lock, /ctl single-socket guard, daemon self-retry semantics)
docs: opencode usage ground truth — response usage always per-turn, usage_update conditional on known context limit (#44 follow-up)
design-cli-tui.md: token accounting rewritten per #44 — dialect survey, response usage, turn_result counting, USD delta
design-cli-tui.md — issue #39 shipped: hnx tui (process model, screen, token accounting, zero-dep renderer, verification)