quitting is a skill

every founder legend is a story about not quitting. the couple more months that saved the company. the cereal boxes. the eighty rejections before the yes. we've built a whole shrine to persistence, and the shrine has a curriculum: hold on, push through, never surrender.

nobody teaches the other skill. which is strange, because every founder i respect is quietly elite at it.

quitting is a skill. knowing what to kill, when to kill it, and how, decided in advance rather than in despair, compounds as hard as persistence does. the point is to move the decision out of your worst moments.

how quitting actually goes wrong

the default quit happens in-episode. you decide the future of a two-year project based on today's mood: last night's sleep, this morning's churn email, that competitor's launch tweet. or the mirror image, you double down in-episode, because quitting today would mean the pain was for nothing.

that second one has a name, sunk cost, but naming it doesn't disarm it, because for a founder it's identity arithmetic. the more you've paid, the more quitting feels like announcing the payment was waste, and you were the one who chose to pay it. i've written about the fog before: early, wrong, and irrelevant feel identical from inside. in that fog, both persistence and quitting are guesses. which means the quality of the guess is decided by the process, and almost nobody has one.

kill criteria

professional poker players solved this problem decades ago, because they face it a hundred times a night: exit decisions made under emotion, with money already in the pot. their solution is boring and it transfers perfectly: decide the exit before the hand.

for a builder that means kill criteria: specific, pre-committed, time-boxed. "if we don't have x paying users by date y, we stop." "if the core retention curve hasn't bent after three full iterations, we stop." written down, when you're calm, fed, and un-desperate.

kill criteria are a gift from your smartest self to your dumbest moment. the you who writes them has perspective and no adrenaline. the you who executes them will have neither, and won't need either, because the decision is already made. all that's left is the courage to keep an appointment.

and no, they're not suicide pacts. you're allowed to override one: consciously, in writing, once, with a new criterion attached. the point is to flip the default from drift to decision. drift is what kills years.

the legends are lying to you (a little)

go back to the persistence shrine and read the stories properly. instagram was a check-in app called burbn that quit everything except the photo filter. twitch was a guy livestreaming his life until they quit that for gaming. slack was the leftover tooling of a game that got killed. youtube quit being a dating site.

the "never quit" canon is actually a canon of elite quitting, one level down. these founders killed tactics ruthlessly, products included, while refusing to kill the mission. that's the real taxonomy: quit tactics fast, quit missions slowly, and never confuse either with quitting your identity, because your identity was never supposed to be in the product anyway.

most bad persistence is a category error: defending a tactic with mission-level loyalty.

the asymmetry nobody prices

here's the cold math that made me respect quitting. persistence has diminishing returns: the difference between iteration three and iteration eight on a dead thesis is usually zero, because the thesis is what's dead. meanwhile the option value of you, the builder, is the single largest asset on your balance sheet, and it's the one every month of zombie-mode burns.

a graceful, early, honest kill refunds almost everything: your energy, your savings runway, your credibility, your appetite. a late, forced, denial-soaked death refunds nothing. you don't just lose the project either way. in the second case you lose the next project too.

quitting well is how you fund the next attempt. i'm on my 19th build. the speed between builds is not despite the quitting. it is the quitting.

the strong objection

"conviction means holding through the dip. your kill criteria would have murdered airbnb in 2008."

honestly: yes. any rule that kills projects will occasionally kill a future winner. there is no criteria set with a zero false-positive rate, and anyone selling you one is lying.

but you are not one company. you are a career of attempts, a portfolio of maybe twenty real swings. across that portfolio, disciplined exits beat always-hold decisively, because always-hold doesn't just cost you the dead years inside each loser. it costs you the winners you never started. the airbnb story cuts both ways: for every team saved by holding on, there's a graveyard of teams who held identical conviction on theses that were simply wrong, and the graveyard doesn't give keynotes.

conviction and kill criteria aren't enemies. a conviction worth having can survive stating what would falsify it. if yours can't, what you have isn't conviction. it's a hostage situation.

keep moving, somewhere

i once wrote that the whole game is to keep moving with ridiculous optimism through the fog. i stand by it. but "keep moving" was never the whole sentence.

keep moving means somewhere. sometimes the strongest move in the fog is to stop walking deeper into it, turn around, and spend your remaining light on a different door.