The keyboard is a bottleneck
We've spent fifty years optimizing for fingers on keys. Every productivity tool, every shortcut assumes you're at a desk, hands on keyboard, eyes on screen.
But your best thinking doesn't happen at your desk.
It happens on walks. In the shower. While cooking dinner. Your mind does its best work when it's free to wander, and in those moments you have no keyboard.
Speech is the original interface
We could speak long before we could type. About 150 words per minute, three times faster than most people type. Speaking doesn't require a special posture or a dedicated device.
We've built an entire computing paradigm that ignores this.
The gap between thought and capture
Every uncaptured idea is a lost idea. Not entirely (you might remember the shape of it later) but the nuance is gone. The specific phrasing. The connections your mind was making right then.
Pull out your phone. Open an app. Find the right note. Start typing with your thumbs. By the time you've done all that, the thought has started to fade. The rich, branching idea gets compressed into a few hurried words.
Voice closes the gap
You speak at the speed of thought. No app to find, no keyboard to fumble with. You just talk.
When transcription happens on your device, in real time, there's no latency and no privacy trade-off. Your words, captured instantly, exactly as you said them.
Voice-first doesn't mean voice-only. Voice is the entry point, the fastest path from thought to text. You edit and organize with whatever interface you want afterward. But capture should be instant.
From capture to action
Raw transcription is messy. Filler words, false starts, tangents. But it contains the real substance of your thinking. The hard part is pulling signal from noise without losing the original intent.
Local AI makes this possible. Models that run on your device, that understand context, that can turn a rambling voice memo into a structured note or a task list. Nothing gets uploaded anywhere.
The pipeline:
- Capture — speak freely, anywhere
- Transcribe — on-device, in real time
- Transform — local AI pulls out structure and meaning
- Route — results land where they belong (notes, tasks, calendar, code)
Everything runs on your hardware.
How we think about it
Local by default. Your voice shouldn't leave your device unless you choose. On-device transcription and AI aren't optional. They're the whole point.
Speed over perfection. A 95% accurate transcript captured in the moment beats a perfect one you never made.
Capture now, organize later. Don't force people to categorize at the moment of inspiration. Let them dump thoughts and sort them out at their desk.
Your tools, your data. Voice is the on-ramp. The off-ramp is whatever you already use: text editor, task manager, email.
What we're building
Talkie is our shot at this. We're building for people who think faster than they type and whose best ideas come at bad times.