I kept losing time explaining my own screen to Claude Code. Screenshot, crop,
paste, then type three paragraphs describing which dropdown I meant. The
context was all right in front of me and I was transcribing it by hand.
Deiko: double-tap Right Option, and it listens while following your cursor.
Hover what you mean while you talk — "this dropdown closes when I scroll, keep
it open". Each thing you point at gets cropped and labelled. Tap Option to
stop, and you get a brief plus the screenshots, which you drag into whatever
agent you use. Works with two languages mixed in one sentence, which is how I
actually talk.
Screenshots and briefs never leave the Mac. Audio is transcribed then deleted;
first 30 minutes use cloud transcription, after that it runs on-device forever.
Things you should know before installing:
- It's signed with a self-signed cert, not an Apple Developer ID, because I
haven't paid the $99. That's why there's an install script: macOS quarantines
the bundle, and the GUI "Open Anyway" misses the Node runtime nested inside,
which then fails silently minutes later mid-transcription. The script is one
`xattr -dr com.apple.quarantine` and a copy to /Applications. Read it first,
it's short.
- The bundle is ~110MB, mostly that Node runtime.
- macOS 14+, Apple Silicon and Intel.
- Pro ($3.99/mo) adds cloud transcription back and your own Sarvam/Groq key.
Everything else is free and always will be.
The thing I'm least sure about is whether pointing-while-talking actually beats
just typing for people who aren't me. If you try it and it felt slower, I'd
genuinely like to hear that.
I kept losing time explaining my own screen to Claude Code. Screenshot, crop, paste, then type three paragraphs describing which dropdown I meant. The context was all right in front of me and I was transcribing it by hand.
Deiko: double-tap Right Option, and it listens while following your cursor. Hover what you mean while you talk — "this dropdown closes when I scroll, keep it open". Each thing you point at gets cropped and labelled. Tap Option to stop, and you get a brief plus the screenshots, which you drag into whatever agent you use. Works with two languages mixed in one sentence, which is how I actually talk.
Screenshots and briefs never leave the Mac. Audio is transcribed then deleted; first 30 minutes use cloud transcription, after that it runs on-device forever.
Things you should know before installing:
- It's signed with a self-signed cert, not an Apple Developer ID, because I haven't paid the $99. That's why there's an install script: macOS quarantines the bundle, and the GUI "Open Anyway" misses the Node runtime nested inside, which then fails silently minutes later mid-transcription. The script is one `xattr -dr com.apple.quarantine` and a copy to /Applications. Read it first, it's short. - The bundle is ~110MB, mostly that Node runtime. - macOS 14+, Apple Silicon and Intel. - Pro ($3.99/mo) adds cloud transcription back and your own Sarvam/Groq key. Everything else is free and always will be.
The thing I'm least sure about is whether pointing-while-talking actually beats just typing for people who aren't me. If you try it and it felt slower, I'd genuinely like to hear that.