What it is
Turns a spoken screen walkthrough into an agent-ready bug report.
What it does: Aligns speech with what you pointed at and selects one picture per point.
Why it belongs in the showcase: Creative multimodal alignment task with a CLI for coding agents.
Caveats: Mac only; privacy trade-off when Jev is on.
Evidence
Public Apache-2.0 repo (2★ at intake) created 2026-10-18. README documents ⌃⌥R recording, on-screen drawing, fn dictation, a `yapsketch` CLI for agents (latest/show/list/frames/path), local speech-to-text by default, and Jev via Opper, OpenRouter or a TypeSafe key. Recordings stay in ~/Movies/YapSketch unless Jev or a cloud speech model is turned on.
Caveats
- macOS only
- Screenshots and words leave the Mac when Jev is enabled
- Early project; build from source
Unofficial case page. Stars / list position observed for heat; do not treat heat as a quality SLA. Review the upstream README before you depend on it.