Bad-take finder.
Four raw takes from my Grok Bot vs Muse vs Dots recording. Each one was transcribed, every sentence was checked by the Decisions API for flubs and restarts, and the bad takes were cut. Open a take to see what was cut and why, and to watch the cut version (720p; a few seconds of my desktop are blurred in 3.2).
4 raw recordings · OpenAI Decisions API (gpt-6-luna) · local whisper
4.8minutes scanned
34.2sseconds taken
$0.0091Decisions API cost
7bad takes cut
Takes
| Recording | Minutes | Scan time | Cost | Cuts | Removed |
|---|---|---|---|---|---|
| 2.5 - 2026-10-06 11-56-55 | 1.22 | 12.3s | $0.0022 | 2 | 3.9s |
| 3.0 - 2026-10-06 13-25-05 | 0.59 | 2.5s | $0.0008 | 1 | 7.5s |
| 3.2 - 2026-10-06 13-27-19 | 1.42 | 7.6s | $0.0032 | 1 | 5.2s |
| 3.8 - 2026-10-06 13-38-47 | 1.52 | 11.8s | $0.0029 | 3 | 8.2s |