HiveCompute wakes up only when your Mac is idle, transcribes 60-second audio fragments locally with Whisper over Metal, and goes quiet again the moment you need your machine back.
See what an idle Mac could make in a month. Then decide whether you want your overnight hours doing nothing or steadily building worker trust and payout history.
The worker is intentionally narrow. It downloads a fragment, transcribes it locally, returns text, and discards the audio.
Whisper runs on your Apple Silicon GPU. The coordinator handles queueing and orchestration, not the actual transcription work.
The worker waits for inactivity and is designed for locked-screen hours, not active daytime usage. You keep control over when it runs.
Electricity draw is modest enough to compare with phone charging rather than full-time desktop compute. That makes idle Mac time economically useful instead of wasteful.
Faster chips clear more 60-second shards per idle hour, which is the main reason monthly estimates vary so much by model.
HiveCompute is designed so your Mac helps with throughput without seeing complete conversations or storing long-lived audio.
Audio is streamed into memory, processed, and discarded after transcription. Nothing about the worker experience is built around inspecting user content.
No single worker hears an entire conversation. Adjacent fragments are assigned to different machines so context stays fragmented by design.
Hidden validation audio verifies accuracy over time. That keeps payout trust tied to real performance, not self-reporting.
Pause the worker, remove the app, or come back to your Mac and let normal use take precedence. The system has to fit your machine, not commandeer it.
Higher-trust workers face less validation overhead and move through payout holdback faster as the coordinator gains confidence.
Highest validation rate and the longest payout hold while the coordinator learns your quality profile.
Clean canary results earn you more production work with fewer interruptions from hidden checks.
Consistent quality makes your Mac more useful to the scheduler and shortens the payout unlock window.
Best-performing workers become dependable fleet backbone machines for high-confidence throughput.
First launch gets you through invitation-code onboarding. The first time your Mac receives real work, it downloads the model it needs, then returns to the background menu-bar flow.
By installing the worker, you agree to the Worker Agreement.