Good morning, NOLA.

Turn editing feedback into a reusable skill

Ad creator Alex Cooper gave Claude Code raw footage and a script, then received a first cut of a Facebook ad after about 20 minutes. He says the result had jolty sections, but he could request changes in plain English instead of opening a video timeline.

The more useful move came next: Cooper converted recurring feedback about pacing, b-roll placement, and other editing choices into a reusable skill. His team also records explanations of creative decisions, turns the transcripts into context, and reviews several agent-produced edits in batches. It is a practitioner’s workflow, not an independent test of Claude’s editing quality.

Make saved videos searchable overnight

Developer Daniel Nwaneri built a local night shift for saved tech videos. Whisper transcribes speech, while Gemma creates searchable notes or reads four frames when a video is silent. The system runs on a CPU-only laptop and pauses when the machine switches to battery power.

The checks are unusually concrete: quoted text must appear word for word in the transcript, and every tool name and number must exist in the source material. The run produced 56 notes from 72 short videos, with eight speech-based notes refused when their quotes failed verification. Nwaneri manually checked ten notes and found nine correct on the first pass.

Also

Share this brief →

Pass it on

Share this brief