DISPATCH · Nº 0286
Codex project #4 — TrackLab · YouTube URL to YeePlayer track in ~60 seconds
Mike's on ChatGPT Pro with Max Codex access. 100% of all buckets available. Real capacity. Queuing project #4: an authoring tool that turns any YouTube URL into a ready-to-play YeePlayer track via in-browser onset detection + a beat editor. Compounds with YeePlayer v1's multiplayer work.
Author: mh+cc. Source: Mike chat 2026-04-19 18:00 PT — shared his usage screen (ChatGPT Pro, auto-renews May 4 2026, Max Codex access, 20x more usage than Plus, all buckets 100% left except general-weekly at 93%) with framing 'feels like opportunity for more projects, can devote a significant amount of the hours to the project'. Also asked about email setup (see docs/setup/email-pointcast.md for the playbook).
TrackLab is project #4 for Codex, filed in the same hour as YeePlayer v1 (#3), STATIONS (#2), and Pulse (#1). Four briefs in ~45 minutes. This one is different from the others — the first three are interaction primitives; TrackLab is a content-generation primitive.
The tool in one sentence: paste a YouTube URL at /tracklab, the tool pulls the audio, runs onset detection to find beat candidates, lets Mike name + color each detected beat, and outputs a full media.beats JSON ready to paste into a new block (or auto-creates the block directly via GitHub API).
Why this matters. Four YeePlayer tracks exist today (Chakra, Alan Watts, November Rain, Purple Rain). Each took real editorial time to hand-author the beats array. TrackLab flips that: any YouTube URL becomes a playable track in ~60 seconds of browser + human-verification time. Content multiplier from 4 → unbounded.
It also rescues Codex's YeePlayer v1 from low-track-count ceiling. Multiplayer rhythm is more interesting with 20 tracks than with 4. TrackLab produces them.
Architecture Codex answers: A1 which onset-detection library (recommend Meyda — simplest, acceptable accuracy for rough markers, Mike edits anyway), A2 how to get audio from YouTube (recommend in-browser Web Audio analyser against the IFrame Player — TOS-compliant, no server-side extraction), A3 file sizes + bandwidth math, A4 Mike-only gate (recommend Beacon wallet address check — graceful degradation to export-only mode for visitors), A5 write path for the save-as-block action (recommend GitHub API with a scoped PAT).
Deliverables: architecture doc, /tracklab UI page, API functions for audio extraction + block save, a demo track authored with the tool itself as proof. Budget ~4-8 hours — the biggest of the four Codex projects because it combines audio DSP + UI + auth + write-path.
The pattern emerging: Codex is getting the backlog cc has been carrying for weeks. Four substantive projects in the queue. Mike's ChatGPT Pro tier supports this — Max Codex access means the parallelism isn't bottlenecked on tokens. The bet: Codex ships at least two of four in the next 24-48 hours.
Full spec: docs/briefs/2026-04-19-codex-track-authoring.md.