1. Computers should feel quieter.
For thirty years we've been adding more — more apps, more tabs, more notifications, more dashboards, more places to look. The result isn't clarity, it's noise. We're trying to build a tool that reduces the surface area you have to manage, not one that adds another lane.
2. The right interface for AI is one key.
Most AI products want to be a destination — open the app, navigate to a tab, type a question, wait, leave. That's a 2010-era model. The right model is ambient: a key on your keyboard, summoned from wherever you are, answers in place, disappears when done. The chrome is the absence of chrome.
3. Local-first or it doesn't ship.
Voice transcription, screen content, and a memory of what you've worked on are the most sensitive things software can touch. If we can run them on your Mac's silicon — and on Apple Silicon, we can — they run on your Mac's silicon. Cloud is an opt-in upgrade, never a default.
4. Mac as the benchmark.
Cross-platform thinking has a cost: you build for the lowest common denominator and everything gets a little duller. We chose macOS first because the Mac sets the visual + interaction bar we want to clear. Windows + Linux versions exist in our heads but they wait their turn.
5. Ship fast, but ship beautiful.
Beautiful is not a polish layer applied at the end. It's a constraint that shapes every decision — what we ship, what we cut, how we organise the team. If something has to be ugly to ship next week, we'd rather ship the week after.
— Daniel, May 2026