A clip machine in an afternoon
To feed the marketing gap, Juan revives Snipy, a parked tool that takes a video plus its transcript and returns a short cut down to the best moments. Built the MVP in four hours (from the 5:30 recording to 11pm), reusing YouTube's own transcript instead of paying a transcription API. The plan: turn each diary's French-progress segment into a short, funnel viewers to Severo, and if people want the tool, package and sell it. The pattern repeats from the Kobo revival: a project abandoned when the tooling wasn't ready becomes a one-afternoon build once the models mature, and this time it exists to solve their own distribution problem.
// trace: where this idea came from
- ↳ video diary @ 23:52 (Snipy, video más transcript, sale un short con lo mejor)
- ↳ video diary @ 24:38 (MVP en cuatro horas)
- ↳ Entry 236-4: Thirty users keep it alive (revivir un proyecto parqueado cuando la herramienta madura)
The marketing bottleneck needs feedstock, and Juan builds the machine to make it. Snipy takes a video and its transcript and returns a short, cut down to the best moments, “masticadito” ▸ 23:52. He’d tried it during the Gearly era and never landed it, but the idea is now trivial: from recording the diary at 5:30 to a working MVP at 11pm, four hours ▸ 24:38.
The build is scrappy in the right places. Rather than pay a transcription API, he uploads the video privately to YouTube and lets YouTube produce the transcript for free ▸ 29:10; an agent reads it with one instruction, pick the best moments and make a video under three minutes ▸ 29:42, and drag-drop the video plus the subtitle file, get a cut back. The first result turns a six-minute diary into a 2:40 short, and it works well enough to use ▸ 35:40.
un proyecto parqueado, hoy un MVP de una tarde →
The move is the same one that revived Kobo: a project shelved when the tooling wasn’t ready becomes a single afternoon’s work once the models mature, and if the shorts pull views, he’ll package Snipy into a product or sell it to run locally ▸ 24:30. What makes this revival different is that it isn’t a side quest, it’s built to solve their own problem, the six minutes of daily French footage nobody has time to edit into something the algorithm will carry. The best tool to reach for is often one you already half-built; the year that passed didn’t waste the idea, it made it cheap…