Omi

The workflow matters here more than the device. Omi records audio through a small clip worn near the collar. The audio goes to the Omi app, and from there I routed it to a webhook on my own server, with Claude Code handling most of the configuration work. A daily cron job runs a summariser against the transcript, extracts decisions and action items, and posts the output somewhere I will actually see it. Worth being clear: none of that is required to use Omi. The app on its own captures and transcribes conversations without any extra work. The friction I ran into is specific to building a developer pipeline, and most buyers will not go near it.

What changed is the gap between having a conversation and having something to do with it. Before Omi, the best available was notes taken during the call, which are partial at best and easy to skip. Now the full transcript exists, and the summariser pulls signal from it without any manual step between the conversation ending and the debrief arriving the next morning. That is the shift worth paying for. It is not the device. It is the automated loop the device makes possible.

The honest limits are worth stating plainly. Omi has trouble picking up whoever you are talking to if they are more than a couple of metres away. That is an expected limitation of the form factor rather than a defect, but it does mean conversations in a noisy space or across a wide table come out partial on one side. The device also disconnects periodically while being worn and requires opening the app to reconnect. Over six months that has been a minor recurring irritant rather than a blocker, but it happens often enough in normal use to mention.