Beeper

Beeper's own search already knows when two chats across networks are the same person, but Merge only joins them when its matcher happens to catch the pair, and misses are silent, so duplicate contacts quietly stay duplicate.

• Traced the gap to a unit-of-organization problem: Beeper treats a chat as the thing it organizes, not a person, even though search already computes the person-level match.
• Sequenced the fix by trust, not effort: reuse the signal search already has, merge on facts like a shared number or email automatically, and only ask once before guessing on a name-and-photo match.
• Drew one line that doesn't move regardless of model quality: never merge on a name alone, since a wrong merge shows one person another person's messages.

Turns Merge from a feature you have to go looking for into a person layer that applies consistently, and makes the case that unlocking it is what justifies paying for the multi-account tiers.

Key gap to instrument first: missed-match rate, since misses are currently invisible to Beeper's own metrics.