The most common meeting problem has no feature attached to it
Somebody joins a call fifteen minutes in. They have missed a decision, two links and one question addressed to them by name. Every video platform on the market responds identically: it shows them a grid of faces and leaves them to infer the rest. The polite ones type "sorry, just joined — what did I miss?" and the meeting stops to re-explain itself. The rest stay quiet and guess.
This is not a rare edge case. It is the single most common complaint about meetings in distributed teams, and it happens to the same person twice a day. It also happens to anyone whose connection dropped and came back, which in practice makes them a late joiner too.
- Cross-timezone teams, where somebody is always arriving into a conversation already underway.
- Consultancies and agencies, where a senior person drops into the last ten minutes of a client call.
- Healthcare, legal and field services — anyone pulled out of a call and back into it.
- Regions where connectivity is variable: a reconnecting participant has missed exactly as much as a late one.
What Bridge shows instead
Bridge treats a meeting as something you can read back, not just a stream that evaporates as it renders. Arrive late and a catch-up card appears on its own, covering seven things: what was said, quoted and attributed to the person who said it; what the room flagged; every link shared while you were away; anything addressed to you by name; the rest of the chat; which signals are still open right now; and who has not spoken yet.
"What the room flagged" is the part that does the work. During the meeting people press Agreed, Differ, Question or Slow down. The card counts how many of them did, with names and timestamps — so you land knowing not just what was proposed but how it went down.
It counts people, not presses
One enthusiastic participant pressing Agreed three times is one person agreeing. It is never consensus. That distinction sounds small and decides whether the feature is trustworthy: a catch-up that inflates agreement is worse than no catch-up, because you act on it.
The card is also extractive rather than generative. Every row in it is something a human deliberately pressed or deliberately said. Unlike an AI summary, it cannot be quietly, plausibly wrong about what happened in the room. Where a written narrative of the discussion is still pending, the card says so rather than papering over the gap.
"But we already have a recording"
A recording is an obligation, not an answer. Nobody watches fourteen minutes of video to work out what they missed before it is their turn to speak. The recording is for the person who was not there at all, days later, and even then it mostly goes unwatched.
Meeting Memory is for the person who is in the room right now and has about three seconds before they need to contribute. Those are different problems and the recording only ever solved the second one.
Why the large platforms do not do this
Not because it is technically hard. Because it requires treating the meeting as a record rather than a stream, and the established architectures discard almost everything the moment it has been rendered to screen. Retrofitting memory onto a pipeline designed to forget is the expensive part, and it is why this gap has stayed open for a decade.
Frequently asked
Does the catch-up card work if I dropped out and rejoined?
Yes. A reconnecting participant has missed the same span as a late joiner, and Bridge treats them the same way.
Is the summary written by AI?
The catch-up card is extractive — quotes, flags, links and mentions that people actually produced. A written narrative of the discussion is a separate capability and the card is explicit about where that gap is rather than filling it with a guess.
