Tripwires

The pattern is Annie Duke's kill criteria, from the poker-and-decision-science record: quitting decisions run systematically late because sunk cost, endowed identity, and status quo bias all pull toward persistence, which means quitting on time will usually feel like quitting too early. Her fix: states and dates, set at deciding time, before the money's spent and the identity's attached.

The operator demonstration is Tobi Lutke's COVID bet. He made a huge capacity bet in 2020 and wrote it down as a falsifiable claim while making it, which meant that when the 2022 numbers landed against it, he could write "I got this wrong" and cut in one motion, instead of spending a year arguing with the evidence the way the sunk-cost version of him would have. The wire didn't make him right. It made him fast at being wrong, which compounds nearly as well.

The mechanism, stated once: a threshold plus a date makes the call self-executing. You don't relitigate daily; you check a number on a date. That's the entire trade, and it converts future doubt from relitigation into observation.

Fixed, three parts: IF [observable state] BY [hard date] THEN [named action].

Two or three wires per decision, rarely more. And where it fits, one upside wire: the state and date that trigger doubling down, because wires fire in both directions and the discipline that kills losers on time is the same discipline that funds winners on time.

Five quality tests, and most homemade wires fail at least two:

1. Observable state. A number or event a stranger could verify. "Under 40 paying customers" passes. "Traction feels weak" fails. "I still dread Mondays on Feb 1" passes, barely: dated, binary, and you'll know.
2. Hard date. A calendar day. "By Q2-ish" is a wire that never trips.
3. Named action. Kill, sell, cut, reopen the full workup. A wire that trips into "we'll discuss it" is a meeting invitation, and the discussion gets held by the warm version of you the wire existed to overrule.
4. No vanity states. Wire the metric that pays, in the currency that matters. Signups when the question is revenue is a vanity wire: kill criteria set on states that flatter.
5. No quiet re-dating. The deadline arrives, the number is close, the date slides a quarter. That's the whole failure mode: relitigation wearing a calendar. A date moves only on genuinely new evidence, in writing, once. Twice is a pattern, and the pattern gets named.

Both directions have teeth. No reopening without a tripped wire or genuinely new evidence: a call reopened on mood gets named as relitigating, out loud. And when a wire trips, the named action executes. Overriding a tripped wire is allowed exactly once per wire, as a new hard call with the full workup, written down, because a hot decision overruling a cold one should have to show its work.

A wire nobody checks is decoration, so every wire gets an owner and a check date. The check takes one minute: tripped or not. Founders who run a Sunday review put the wires there, which is why the system's weekly cadence exists at all.

Founder adds an outbound sales motion to a $70k MRR SaaS, three months of budget committed. Wires set the same day, cold: IF under 8 qualified demos a week BY Dec 1 THEN kill outbound, back to founder-led sales. IF blended CAC over $2,400 BY Jan 15 THEN kill. UPSIDE: IF 3 closed deals over $10k ACV BY Feb 1 THEN hire a second SDR.

Mid-November is ugly, and after one bad week he wants to kill the whole thing. The sheet says no wire tripped. The doubt gets logged as weather, and he goes back to work. December 1: nine demos a week. Holds. January 15: CAC lands at $3,100. The wire trips, outbound dies that week, no meeting, no agonizing. The agonizing happened in October, when he was cold and could think. That's the entire product: the argument happens once, at the best possible temperature, and the future just reads the gauge.