Should long-running agents be priced on the final outcome?

THE SHORT ANSWER

No. Keep the finish line as the renewal number and price a leg. David Haber at a16z argues that long-running agents can own months-long journeys and be paid on the outcome, with a settled plaintiff case as the clean unit. James Brown's Stripe write-up of Monetize 2026, the day before, reports that Fin's single-outcome meter (per resolved conversation) stopped fitting once the agent did a second thing and became multiple outcomes under a credit pool. I priced one outcome, a resolved ticket with a 48-hour definition. It held for 98% of cases; the other 2% became more than 200 contested cases a week and a dispute team that grew from 4 to 14. A relay has a definition at every handoff, and the customer disputes the leg they believe the agent did not run. Price the leg whose definition they will sign, with the dispute window and arbitration in the contract, and report the settlement for renewal.

The product argument in Haber's piece is good: a relay leaks value at every handoff, and an agent that holds the whole record and never stops calling the witness closes leaks no human team can. The pricing argument at the end of it assumes "identifiable" and "agreed" are the same word. They weren't on one ticket. They won't be on five handoffs.

The handoff table, the three questions, and the Stripe field report are in Kill the Finish-Line Price. A Relay Has a Definition Per Handoff. The dispute numbers are in The Output Gets Disputed Too.

SOURCES

THE LONG VERSION

RELATED ANSWERS

Last reviewed 2026-10-14 · 1 min read