You're Writing to a Stranger
Notes are addressed to your future self. But research shows we experience future-us like strangers — and we write accordingly, to our cost.
The person who wrote your notes from eighteen months ago was you. But reading them can feel like reading someone else's shorthand — abbreviations whose expansion you've lost, references to conversations you can barely place, conclusions without the premises that made them obvious.
The usual explanation is that memory fades. That's partly true. But there's a sharper account.
Your future self is someone you barely know
In 2011, Hal Hershfield at NYU published a series of studies on what he called temporal self-continuity. The premise was disarming: when people think about their future selves, how strongly do they identify with that future person? The answer, consistently, was not very much.
In fMRI experiments, Hershfield's team found that the neural signature of thinking about your future self resembles the signature of thinking about a stranger more than the signature of thinking about your current self. You and your future self share a body and a name, but the psychological identification is weak — closer to how you think about a fictional character than how you think about someone standing in the same room.
This has consequences for financial decisions (people underinvest in retirement because future-them feels like someone else's problem), and it has consequences for note-taking, though nobody writes that part up.
When you write a note, you're implicitly addressing someone. Most of the time you write as if that person is you — the exact same you, with the same context, the same state of mind, the same abbreviations decoded. You write "talk to M about the handoff" without specifying which M, which handoff, which context. You write "switched to approach B" without recording why B, what A was, what made them distinguishable.
You're not being careless. You're just miscalibrating for who will read this.
What notes assume
Every note makes assumptions about its reader. The shallower the note, the more it assumes. "Follow up re: thing from yesterday" expects your reader to remember yesterday, know what the thing is, and understand what follow-up means in this specific context. You, in the next twenty minutes, can decode that. You, in eight months, probably can't.
Software engineering has a partial answer to this. In 2011, Michael Nygard popularized the Architecture Decision Record — a lightweight format for documenting technical choices. The format has four fields: context, decision, status, consequences. The context field is the part that matters. You write down the situation you were in, the constraints you were under, the alternatives you actually considered. Not just what you decided, but why that decision existed at all.
Teams adopted ADRs because they're writing notes that engineers two or three years from now will need to interpret. Those engineers will effectively be strangers — different mental state, different context, sometimes literally different people. The notes have to assume less.
The trouble is that writing for a stranger is genuinely more work. Hershfield found you can improve people's retirement savings decisions by showing them age-progressed photos of themselves — making future-them feel more like a continuous, real person who will actually experience the consequences of today's choices. There's no equivalent trick for note-taking. No age-progressed face to make you write a slightly fuller sentence before closing the tab.
Where context lives when notes can't hold it
A standalone note is a closed system. It holds exactly what you typed. If you wrote tersely, it's terse; there's no surrounding structure that can expand it.
A knowledge base with rich structured records is different. If the note was written inside a project with a record of your goals, your constraints, the people involved, the open questions — then a sparse note sits inside a context that can partially decode it. "Switch to approach B" means more when the same workspace shows what project you were working on, who you'd spoken to that week, what alternatives were on the table. The note stays sparse. The surrounding structure isn't.
An AI with access to that context can do something your memory can't: brief your future self. Not just retrieve the note, but situate it — what was active when you wrote it, what decisions were live, what you were trying to figure out. The note doesn't need to be more detailed if the environment around it is rich enough to provide what the note left out.
That's the honest version of AI memory. Not a perfect transcript of everything you've ever thought. A structured enough context that your present self's sparse communication to your future self doesn't get completely lost in translation.
You're still writing to a stranger. You just have a better introduction.
Asgeir Albretsen is the founder of Harbor.