About

A tool that tells you when it cannot do something.

This page exists mostly to say what we will not tell you.

The idea

A technical interview is half an hour where you are asked to think out loud, in public, at speed, about something you may last have touched two years ago. The gap that ruins people is not knowledge — it is the thirty seconds between hearing a question and finding the shape of an answer.

So the product does one thing: it listens to the call, and on one keystroke it turns the last thirty seconds of what was said into a worked answer. No copy-paste, no retyping the question into another window, no alt-tab a camera can see. Success is the answer arriving before you have finished reading the problem.

What we refuse to claim

This is a category that invites inflation, sold to people professionally trained to detect it. So there is a rule in this repository: every number on this site has to trace to something real in the source, and the ones that cannot are deleted rather than rounded.

  • No customers, testimonials, case studies, press quotes, user counts or funding. There are none, so there are none on the site.
  • No time-to-answer figure. It has never been measured, and an unmeasured latency claim is the easiest lie in software.
  • No language count for transcription. Coverage is whatever the speech engine supports, which is not a number we can put our name to.
  • No benchmarks against other tools, because we have not run any.

The same rule cuts the other way. Session history is capped at the 50 most recent rounds, and screenshots are kept as thumbnails rather than full frames. Those limits are stated on the pages where the matching capability is stated, not buried in a footnote.

How it is built

A desktop application, not a browser extension and not a meeting-bot that joins your call as a participant. It captures your microphone and the system-audio loopback as two separate channels, which is how every transcript line gets attributed to whoever said it, and it is why it does not care whether you are on Zoom, Meet, Teams or a browser tab.

Sessions are written to your own disk. 2 speech engines are supported behind the scenes. There are 18 global hotkeys and all of them can be rebound. The window floats above whatever is in front of it, with adjustable opacity and click-through, because the interface competing for attention during an interview is itself a failure.

Who

who builds this — names, or a sentence about the team. Reachable on thecontact page, which routes to a person rather than a ticketing system.