AZNA

I recently earned my 50-hour Yoga Sculpt teaching certification, and one pain point kept surfacing in training and in conversations with working instructors. The slowest part of preparing a class is syncing the playlist to the sequence. Musicality can make or break a class, and a class where the music is out of step with the movement falls flat.

So I built myself a Claude Code skill that plans the class and times every block to the actual tracks. A few real classes later, the market was clear: most instructors teach as a passion project or a side job alongside a full-time career, and prep time is their scarcest resource.

AZNA gives instructors that prep time back while keeping every class personal to them: their sequence, their music, their voice, assembled faster. It is fully cross-platform: a Next.js web app, a native iOS app, and an Apple Watch companion that runs the live class clock with haptic countdowns, HealthKit workout tracking, and two-way state sync with the phone. Spotify connects through OAuth, a BPM engine times each block to real tracks, and the finished playlist lands in the instructor's own account. An LLM cue writer drafts coaching cues for every block, streamed over server-sent events, validated in code, and backed by a deterministic fallback so a bad model answer can never break a build.

The engineering underneath is production-grade: 37 rate-limited API routes, 22 database tables carrying 45 row-level security policies, server-enforced quotas and entitlements behind the trial, free, and $9.99 Pro tiers, and a 63-file test suite with an LLM-as-judge eval harness, run in CI on every push. Six days and 110 solo commits took it from first commit to a fully branded product staged across all three platforms, built on the deterministic timing engine from my original Python skill, ported to TypeScript and tested to numeric parity. The public App Store launch is the remaining step.