You don't really decide whether to run tomorrow. A memory decides — your memory of the last run. And that memory is not a recording. It's a heavily edited highlight reel, assembled by a narrator who discards almost everything that happened and keeps two frames: the worst moment and the final one. Which means you can feel fine for twenty-five minutes of a thirty-minute workout, finish it gasping and demoralized, and walk away carrying a memory that whispers, when the alarm goes off tomorrow, that was miserable — skip it. The workout didn't fail you. The ending did.

The two selves that never meet

In the 1990s, the psychologist Daniel Kahneman began drawing a distinction that would work its way into nearly every corner of behavioral science: there is the experiencing self, the one living your life moment by moment, and the remembering self, the one who keeps the records and makes the decisions. They are not the same person, and they frequently disagree.

The cleanest demonstration is almost comically simple. In a 1993 study, Kahneman and his colleagues asked people to hold a hand in painfully cold water for sixty seconds. In a second trial, the same people held a hand in the same cold water for sixty seconds — followed by thirty additional seconds during which the water was warmed slightly, from painful to merely unpleasant. Objectively, the second trial contained strictly more discomfort: everything the first one had, plus half a minute extra. Then the researchers asked which trial people would rather repeat. Most chose the longer one.

Read that again. People voluntarily signed up for more total suffering, because the suffering ended better. The remembering self doesn't add up pain the way an accountant would. It takes something closer to an average of the most intense moment and the final moment — the peak and the end — and files that summary away as "what the experience was like." Almost everything else, including how long the experience lasted, falls out of the ledger. Researchers call that omission duration neglect, and Kahneman and the physician Donald Redelmeier found the same pattern in a very different setting: patients' memories of a painful medical procedure tracked the worst moment and the final moments, not how long the procedure actually went on.

This is the peak-end rule, and it is quietly running your habit life.

Duration neglect: why effort doesn't buy credit

Here's the part that stings if you've ever ground through a long, virtuous session of anything. The remembering self gives you almost no credit for duration. A ninety-minute study session that ends in fog and frustration gets remembered as worse than a focused forty-minute one that ends on a solved problem — even though you did more than twice the work.

That feels unjust, and it is. But it has a design logic. Memory's job isn't to archive your life; it's to compress your life into something usable for future decisions. A summary built from "how bad did it get?" and "how did it end?" is a cheap, fast heuristic that serves you well most of the time. The trouble starts when the thing being summarized is a habit you're trying to build — because a habit lives or dies on one question, asked over and over: do I want to do that again?

And who answers that question? Not the self who actually lived through the session. The narrator. The editor. The remembering self, holding its two frames.

Your habits run on the highlight reel

Look at how most people structure their hard habits and you'll notice something perverse: we routinely arrange them so that the worst moment is the ending.

The runner finishes with a lung-burning push up the final hill. The new meditator spends the last two minutes fighting a wandering mind and closes the session with a silent verdict of I'm bad at this. The writer stops when the words dry up — by definition, at the exact moment the work feels most hopeless. The person doing a monthly budget review quits at the most depressing line item, because that's where the will to continue died. In every case, the session ends at its low point, and the peak-end rule dutifully stamps the entire experience with that low.

Then we wonder why tomorrow's version of us doesn't want to come back. We blame discipline, motivation, character. But the aversion isn't a character flaw — it's an accurate response to a corrupted record. Your brain is steering you away from something its own summary sincerely reports was awful. The experiencing self may have had a perfectly decent time for 90 percent of the session. The experiencing self doesn't get a vote.

There's encouraging evidence that this is fixable, and not just in theory. In a 2016 study, Zachary Zenko, Panteleimon Ekkekakis, and Dan Ariely had people complete cycling workouts with the same total work but different shapes: for some the intensity ramped up over the session, for others it ramped down. The ramp-down group — identical effort, gentler ending — remembered the workout as more pleasant and forecast more enjoyment of future exercise. Nothing about the workout changed except its trajectory. The ending rewrote the memory, and the memory is what shows up next time you're deciding whether to lace up.

Ending is a skill, not an afterthought

Most habit advice obsesses over beginnings — cues, triggers, getting started. Fair enough; you can't finish what you never start. But the peak-end rule says the last two minutes of a habit may matter more than the first twenty, because the last two minutes are disproportionately what your future self will consult.

That reframe changes what a "good session" even means. A good session is not the one where you squeezed out maximum volume until you hit the wall. A good session is one that ends while the ending can still be good — because that session produces a memory that pulls you back, and the session that pulls you back beats the heroic one that doesn't. Every week. Compounding.

This is also why "leave something in the tank" is ancient athletic wisdom, and why some writers famously stop mid-scene while they still know exactly what happens next. They aren't being precious. They're manufacturing tomorrow's appetite.

Your next moves

  • Ramp down, don't gut out. Restructure the final five minutes of your hard habit to be the easiest part: an easy jog after the hill, light stretching after the heavy lifts, rereading a paragraph you like after writing a difficult one. Same total work, different shape, different memory.
  • Stop one rep early — today. End your next session at the moment it's still going tolerably well: one rep, one paragraph, one page before you'd naturally collapse. It will feel like leaving money on the table. You're actually buying tomorrow's session.
  • Install a 30-second closing ritual. Before you walk away, name one specific thing that went fine — "I showed up on four hours of sleep," "page three has one good sentence." You are deliberately authoring the final frame the peak-end rule will use.
  • Move the worst part to the middle. If your routine contains a part you dread — the burpees, the scariest tab of the spreadsheet — schedule it mid-session, never last. The peak still counts against you, but at least it won't also be the end.
  • Audit one aversion this week. Pick a habit you've been avoiding and ask: how did the last session actually end? If the honest answer is "badly," you may not have a motivation problem at all — just a bad final frame. Rerun the habit once with a designed ending and notice what shifts.

The version of you that decides

The self that decides whether you show up tomorrow was never at today's session. It only has the summary — two frames, peak and end — and it will trust that summary completely. This is exactly why small steps beat heroic ones: a habit small enough to finish cleanly ends on completion, every single time, and quietly builds a library of memories that say that was fine, do it again. That's the idea Cadence is built around — habits scaled down until ending well is the default, tracked so you can watch the pull toward them grow. If you want help designing habits your remembering self will actually vote for, you can start at cadence.lumenlabs.works.