solaoi

Intercat is a macOS timeline editor that lets people direct multiple AI agents while their work is running. As agents work in parallel, the person supervising them has to track what each one is doing, revise upcoming instructions, decide which actions to allow, and determine which results to keep. Intercat brings those decisions onto one shared timeline. The central interaction is editing work in time. Each agent has a track. Type an instruction and the local Editor helps place it after the appropriate worker's current turn. The pending instruction appears as a dashed clip that can be cut before execution. An approval request stays with the affected track, so another agent can continue while the user reviews the command and its reason. Completed steps retain their history, tool output, and file changes. The Commit sheet then makes the included steps and external effects visible before the user accepts the result. The demo makes these behaviors directly observable. Real Muse Code workers run on a disposable software project. A pending instruction is cut while active work continues. A requested git push is held before execution, and the remote is checked for changes. The worker host is killed, the same session reconnects with its history, and the resulting work is reviewed and committed. In that recorded run, Glimmer's instruction-placement inference took 0.50 seconds, and recovery after the worker-host kill took 4.1 seconds. One of 14 worker approval requests required a human decision. These are measurements from this run. The public repository includes source, setup instructions, tests, and a checklist documenting 60 acceptance criteria verified through automated checks and live-worker testing. All features shown were built during the hackathon, including the timeline UI, local Editor integration, Rust orchestration daemon, per-track approvals, persistent history and recovery, and Git integration. The product addresses coordination across AI tasks. The demo uses a software project to make the actions and resulting changes easy to verify.