Milestones

List view

• Tests & tooling

  Test suites, e2e, bench and CI gates give verdicts about the code, never about the box: flakes, leaked daemons, bench cells and canaries that record the truth, and gates with no holes.

  No due date

  •0/8 issues closed
  0% complete8 open0 closed

• Chat & connections

  New and sharper chat surface work plus connecting and switching the model accounts codeaf runs on: side chats and forks, transcript grammar (lens), capture, wait clock, questions, home/task rows, voice — and named, switchable model connections (OpenRouter and custom: Ollama, vLLM, self-hosted, OpenAI-compatible) added and managed from /connect and /settings. Features and polish — not release blockers.

  No due date

  •6/16 issues closed
  37% complete10 open6 closed

• Reliable agent

  Tasks, spend/routing, and headless endings behave as claimed — checks, worktrees, receipts, model ladders, and spend that match reality.

  No due date

  •17/51 issues closed
  33% complete34 open17 closed

• Remote & multi-attach

  Reach and steer a conversation from another terminal over a real link — multi-attach, pairing, and remote doors that tell the truth. A capability wave, not a ship gate.

  No due date

  •10/15 issues closed
  66% complete5 open10 closed