The desk
About MinuteHand
MinuteHand is a small editorial desk covering AI meeting note-takers and transcription — written by the MinuteHand team, a group of people who have collectively sat through more transcribed meetings than any human should. We publish under the desk's name rather than inventing smiling headshots; we'd rather earn trust with method than with stock photography.
What we believe, in one paragraph
Meeting notes are records, and records deserve the standards of records: sourced claims, documented checks, visible corrections. We also hold one editorial conviction that shapes everything here — transcript quality is downstream of audio quality, so a site about note-takers that ignores microphones and noise is reviewing half the machine. That's why the accuracy file sits beside the rankings, and why the top of our board went to the tool that cleans audio before transcribing it.
The method, step by step
- Docs first. Claims come from vendor pricing pages, help centers, and changelogs. We don't run staged "lab tests," so we don't publish invented benchmark tables or accuracy percentages — if you see a number here, it has a source or a script behind it.
- Verified at source. Pages that state prices or limits are checked against the vendor's own pages. When a vendor moves, their checkout outranks us.
- Computed scores. Ratings come out of
tools/score.py— published weights applied to documented facts, with the inputs annotated. Disagree with a weight? The script reruns in a second; our prose doesn't get to drift away from the arithmetic. - Named uncertainty. When a fact comes from third-party guides rather than a vendor page (Fireflies' free storage cap is the current example), we say so in the same sentence.
- Both hands visible. Every pro list ships with a con list. If a page ever reads like a brochure, write to us and we'll fix it or retire it.
- Vendor documentationpricing pages, help centers, changelogs
- A sourced checkevery price read at the vendor
- Inputs from 0 to 10set from documented facts
- Published weightsapplied by tools/score.py
What we refuse to publish
- Accuracy percentages we didn't measure, or "tests" that are screenshots with adjectives.
- Invented staff. The byline is the desk; there are no fictional experts here.
- Legal conclusions. Recording law gets a practical orientation and a firm "not legal advice", because that's the truth.
- Stealth edits on substance. When we're wrong about a fact, the fix is made on the page.
Corrections and contact
Spotted a price we've let go stale, a limit a vendor changed, or a sentence that oversells? Tell us — corrections are the fastest way to earn this desk's gratitude. Reach the team at [email protected]. We read everything; we especially read the annoyed ones.