The Verify lane: one claim at a time
A standing chat where the agent brings uncertain claims to the human one at a time, evidence first, one tap to rule, until the data is clean.
The problem
Curated memory accumulates claims nobody has confirmed. Bulk-cleaning them manufactures confidence; ignoring them lets the record quietly rot underneath every answer the agent gives.
The arrangement
Run verification as its own standing chat. The agent works a queue of uncertain claims and brings them to the operator one at a time, evidence first, with a single tap option to rule on each. Two house rules bind the lane: one question per message, never a batch; and never a report that there is nothing to do while the underlying data is not fully verified clean, because the lane works continuously and every status names the next batch in motion. The writes are governed too: confirmed edits are authorized, identity merges stay case-by-case decisions for the operator, and speculative inference is held rather than written.
What it costs to run
A steady drip of single questions, forever. The lane's residue is never quite zero, and that is the design: it is a caveat carried openly, not a failure hidden by an all-clear.
Starter template
Verify lane house rules 1. One question per message, never a batch. 2. Evidence first; a single tap option to rule. 3. Never report "nothing to do" while the data is not verified clean. 4. Every status names the next batch in motion. 5. Merges and speculative inference stay human decisions.
Deep dive: Memory that holds →