An independent register of AI capture devices and the agents that turn transcripts into a usable wiki. Everything here is graded on one question: does audio get from your mouth to an AI with zero manual steps?
Built because the vendor comparisons all measure the wrong thing. Battery life and microphone count do not matter if the transcript dies inside an app you cannot reach.
The grade is not an opinion about quality. It answers whether a transcript can reach an AI agent unattended. A beautifully built recorder that traps its output in a phone app gets a low grade here and might still be the right purchase for someone else.
Strong meets four of five. Viable means a real path exists and needs plumbing or a tier upgrade. Falls short means capture works and handoff does not. Not applicable means it was never a capture pipeline.
Deep means web-verified by hand. Light means the category answer already decides it, so per-device research would change nothing. Source only means the claim came from a starting list and has not been checked. Open question names the one specific thing that would settle a row.
This register started from a list that turned out to contain products that do not exist. Corrections are as welcome as additions.
Our cadence, stated plainly. Submissions publish immediately, badged as unverified. We clear a batch of ten on a weekly sweep and promote the real ones into the graded register. We do not promise to review every submission on arrival, because that is a promise that breaks.