I'll remember this

"I'll remember this."

You've said it. Probably while driving, or walking between meetings, or halfway through an article you were already forgetting. You had a thought — a connection, a name, a decision — and you made a prediction: this one I'll hold. Then it was gone.

The prediction isn't a failure of willpower. It's a failure mode built into how prospective memory works.

Gilles Einstein and Mark McDaniel spent decades studying prospective memory — the kind that involves remembering to do something in the future, rather than recalling something from the past. Their consistent finding: we're good at forming prospective intentions but terrible at judging their reliability. The confidence we feel about a future memory has almost no relationship to whether the memory will actually be there when we need it. You feel certain you'll remember. You don't. There's no correlation.

The "feeling of knowing" — a related phenomenon studied by J.T. Hart in 1965 — is also deceptive: the sense that information is in there somewhere, just out of reach, often precedes a genuine inability to retrieve it. You feel close to something that isn't actually accessible.

Together these suggest something uncomfortable: the moment you think "I'll remember this," your subjective confidence peaks while your actual odds of retrieval remain flat. You're most certain at exactly the wrong time.

The invisible cost

The frustrating thing about failed prospective memory isn't forgetting. It's knowing you knew.

Forgetting something you never captured is clean. You reach for it, find nothing, move on. But reaching for something you're certain you once knew — something you specifically decided was worth holding — produces a particular cognitive friction. You remember the decision to remember. You just don't have what you decided to remember.

This is the hidden cost the "I'll remember this" prediction produces when it fails. Not just missing information. The certainty that it should be there, which makes absence feel like malfunction rather than ordinary forgetting.

Why capture feels optional

Part of why we make bad predictions here is that the cost is deferred. At the moment of "I'll remember this," there's no penalty for being wrong. The penalty arrives later, in a different context, usually at exactly the wrong time.

Information tools that make capture easier change this a little. But not as much as we'd expect. The barrier to capture isn't usually friction with the app. It's the threshold decision: is this worth capturing at all? That decision is made fast, in the heat of the original context, without the future self's perspective.

The things that don't get captured aren't, on average, things people judged unimportant. They're things people judged memorable. The act of thinking "I'll remember this" actively reduces the likelihood you'll capture it, because it substitutes a prediction for an action. You feel like you've done something about preservation. You haven't.

Linda Henkel's 2014 photo-taking research found something structurally similar. When people photographed an object rather than studying it, their later recall dropped. The act of taking the photo substituted for the cognitive work of encoding. "I'll remember this" might work the same way. The prediction provides a false sense of having handled it.

What purposeful capture changes

The standard response to this problem is frictionless capture: fast notes, voice memos, quick-add shortcuts. And lower friction helps at the margins. But it doesn't address the threshold decision — the moment you have to decide whether this particular thing is worth doing anything about.

What changes the threshold is not speed but purpose. When you know that capturing something means naming what kind of thing it is — a person, a decision, a preference, an open question — the act of capture carries enough weight to compete with the shortcut. You're not just throwing something into a pile. You're categorizing. And that's a different cognitive act than prediction.

Typed entities work here not because they're faster. They're sometimes slower. They work because they transform the capture question from "is this worth remembering?" to "what is this?" That second question is easier to answer — it's about observation, not forecasting — and answering it is most of the cognitive work that makes knowledge useful later.

The goal isn't to capture everything. It's to stop trusting a prediction that prospective memory research suggests fails far more often than it succeeds.

You've made the prediction before. You'll make it again. The only thing you can actually change is what you do in the half-second before you make it.

---

Asgeir Albretsen is the founder of Harbor.