Show HN: Subvocal – Talk to your computer without making noise
- ID
- 42135
- Status
- summarized
- Published
- 03 Nov 2026, 1:16 AM
- Fetched
- 03 Nov 2026, 3:34 AM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://subvocal.company/
- Source URL
- https://hnrss.org/best
Summary
- Score
- 4.5
- Created
- 03 Nov 2026, 3:35 AM
- Tags
- Audience
- developersvibe_codersai_agent_users
What happened
Subvocal is a Y Combinator-backed hardware product that claims to decode "almost-invisible" silent mouthing into text at 200+ WPM, aimed at messaging, notes, and dictating prompts to coding agents. The page explicitly states it uses no cameras and does no lip-reading or thought-reading, and pitches scenarios like dictating to Claude Code at a shared desk. It is a preorder landing page: no price, ship date, accuracy data, or independent testing is given.
Why it matters
There is nothing to decide yet — no price, no ship date, and no measured accuracy, only a 200+ WPM claim from the vendor itself. The one concrete thing to note is the framing: the demo use case is dictating prompts to an agent like Claude Code, which means the product's usefulness hinges on accuracy for code identifiers and punctuation-heavy prompts, not conversational speech — a bar the page does not address.
Discussion angle
If silent dictation really hits 200+ WPM, does it hold up on the stuff developers actually say — `useEffect`, snake_case names, brackets, and quote marks — or is the 200 WPM figure only plausible for plain prose?