The old stitch (boot-jeffrey.mp4 + demo-viennese.mkv) double-counted the boot greeting — boot-jeffrey's TTS played over demo-viennese's own boot, causing the user-reported "notepat sound doesn't match when it appears". Drop boot-jeffrey.mp4; demo-viennese already has its own 3-second boot greeting with the matching "hi @jeffrey · enjoy Los Angeles!" TTS. New 21s sequence: 0.00 – 3.00 boot greeting (demo-viennese intro, has own TTS) 3.00 – 17.75 notepat playing the Viennese waltz 17.75 – 19.75 prompt piece — "off" typed character by character (sim) 19.75 – 21.25 shutdown strobe with "bye @jeffrey" (sim) Narration retimed and pruned to match: [0:04] notepat - the default piece. [0:07] Thirty-two voices, 192 kilohertz. [0:11] Chord colors track what you play. [0:14] A Viennese waltz, played live. [0:18] Back to the prompt. Type 'off'. [0:20] --silence-- Key subtitle constraint enforced: no narration during the 0-3s boot (intro has its own TTS) or the 19.75-21.25s shutdown (ac-native plays its own "bye @handle" TTS + shutdown chime). Subtitles only during the instrument-showcase middle and the "type off" moment. Added a prompt-off simulation (tools-adjacent Python, not committed separately — one-off render in /tmp) that draws the AC Native prompt piece with "> off" appearing char-by-char, blinking acpink cursor, legend ("type a piece name - off to power down"). 2 seconds, matches the boot's cream-on-warm-white palette. Sequence chain chip updated to reflect the real flow, with an honest byline noting prompt + shutdown are simulated until ac-native rebuilds with the new draw_shutdown_anim() hook and captures the hardware version. Cachebuster bumped v=3 → v=4. New videos live on CDN. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>