WakeProof

WakeProof is an iOS alarm that uses Claude Opus 4.7 vision + reasoning as a self-commitment device, not just a feature. Capture a baseline photo during onboarding; when the alarm rings, the only way to dismiss it is to take a live photo that Opus 4.7 verifies against the baseline. Unlike pixel-match "photo missions" in alarms like Alarmy (unchanged since 2014), Opus 4.7 reasons about authenticity. It catches printed photos, screen displays held to the camera, half-closed eyes, missing micro-expressions — every shortcut a sleep-deprived user might attempt at 6 AM. The same Opus 4.7 call does three jobs in one model: vision (face + morning state recognition), reasoning (anti-spoof detection with traceable judgment), and insight (writes a personalized weekly reflection from accumulated wake history). A naive build would stitch three narrower models; with Opus 4.7 it's one call. Around the verification core, three retention hooks reinforce the commitment loop, each tied to a specific behavioural-economics signal: a Streak Calendar (continuity visual feedback) gamifying consecutive verified mornings, an Investment Dashboard (accumulated-data loss aversion) surfacing the user's growing trust profile, and an opt-in Share Card (social reward + autonomy preservation) that turns a streak into a postable artefact without the product ever auto-posting. None of these use Opus 4.7 — they're what make the verification meaningful beyond the moment it happens, turning a one-time demo into something that compounds. The result is an alarm that becomes a contract you can't unsign. The Apple Clock alarm can be muted in two seconds. WakeProof can't — not without proving you're actually upright, in the right room, with your eyes open. Built solo in 5 days. Local-first SwiftData. No backend except the Opus 4.7 verification call. Open source under MIT.