Scouted · October 13, 2026
fast voice control for Mac tasks
A clever local-first voice router, but it's a thin wrapper around Claude CLI with no moat, no billing, and a crowded macOS assistant market.
Why now?
The Show HN post drew 107 points and 23 comments, indicating immediate interest in a hold-to-talk Mac assistant that routes simple tasks locally instead of through an LLM. The builder's own framing — simple tasks like opening apps or Spotify playlists finish in under a second without touching AI — reflects a current shift toward local-first, low-latency voice control as LLM API costs and latency become friction points for always-on assistants.
The gap
Existing Mac assistants (Raycast, Alfred, Apple Shortcuts) are keyboard- or GUI-first, and most voice assistants send every request to a cloud model. Atticus's differentiator is a local router that handles common commands without AI, reserving Claude CLI for complex tasks. The HN response suggests this hybrid is not well served, though the signals are limited to one post.
Main competitor
Raycast, Alfred, and Apple Shortcuts — established Mac launchers with large user bases, though not hold-to-talk voice-first and generally without a local/AI task router.
Execution plan
- Ship a signed, notarized macOS build with a one-line install (Homebrew cask) so users can try it without cloning a repo.
- Add a configurable local command router with a visible list of supported intents (apps, Spotify, web search) and a fallback to Claude CLI only when no local match exists.
- Instrument latency and cost per request locally so users can see the local-vs-AI split, then publish those numbers as the core marketing claim.
- Open-source the router and intent definitions to attract contributors and reduce dependence on a single maintainer, while keeping the app distribution closed if monetization is pursued.
- Collect structured feedback from the 23 HN commenters and iterate on the top requested intents before broadening scope.
Monetization
No pricing or billing signals exist in the data. A plausible path is a paid Mac app license for the signed build, with the open-source router as a free tier. Willingness to pay is not yet proven — one HN post shows interest, not demand.