Notes Apps Are Lying to You About Search

There's a specific feeling that note-takers know well. You wrote something down — you're sure of it. A decision, an idea, something someone told you that felt worth keeping. You open the search bar and type what you remember. The results come back thin, or wrong, or empty. You try three more phrasings. Nothing. You close the app and try to remember on your own, which is exactly what you were trying to avoid.

Psychologists call the human version of this "tip of the tongue" — knowing something is in memory while being unable to retrieve it. It happens to most people one to four times per week. What nobody mentions is that note-taking apps reliably produce the same feeling, in an app full of things you actually wrote down.

This is worth sitting with. You took the time to write it. The app stored it faithfully. And you still can't find it. That's not user error. That's a broken promise wearing a search bar.

The promise was recall

Evernote launched in 2008 with a tagline: "Remember Everything." It was not saying "store everything." It was making a claim about retrieval — that writing things into Evernote meant you wouldn't have to rely on your brain to hold them. The note would be there when you needed it. You would remember, via the app.

That promise was never really fulfilled, and the reason is almost embarrassingly simple: keyword search requires you to remember the words you used.

Type "budget" and it won't find a note where you wrote "cost estimate." Type "hesitant" and it misses the document where you wrote "wasn't sure." Type "Q3 launch" and you'll never surface the note that said "summer release." Any gap between how you wrote something — weeks ago, in a different mood, with different vocabulary — and how you're searching for it now becomes an invisible wall. The note is there. The search returns nothing.

This is not a failure of implementation. It's a category error. Search finds strings. Retrieval finds meaning. The app ships search and calls it retrieval. Most people never notice the distinction until the moment they need the thing they wrote down.

Why capture became the product

Capture is an easy problem. It's fast to build, satisfying to use, and it creates a measurable artifact: a note exists now that didn't before. You can see it. You feel organized. The friction of writing things down has been engineered nearly to zero — keyboard shortcuts, quick-capture panes, web clippers, voice memos.

Retrieval is the hard problem. It's hard to build, hard to demonstrate, and its success is measured by an absence: the frustrated search that doesn't happen, because what you needed was surfaced before you had to ask for it.

The result is that almost every notes app is excellent at capture and mediocre to poor at retrieval. The search bar exists because you need something in the top-right corner. It works for simple lookups. It creates the impression of findability.

The impression is largely theatrical.

What it would mean to actually solve it

When people talk about search in notes apps, they usually mean one thing: a box you type into that returns documents. What would actually solve the recall problem is different — it's closer to a queryable knowledge layer that doesn't require you to remember your own vocabulary.

A note where you wrote "Kristian mentioned he prefers shorter proposals" is, on its face, a string. But "Kristian's preference about proposals" is a fact — a typed entity, attached to a person record, retrievable through a structured query rather than a string match. The difference matters when you're on a call and need to quickly remember what someone told you. You're not going to remember you used the word "shorter." You are going to want to know what Kristian prefers.

Semantic search helps. Converting documents to vectors and searching by meaning rather than exact strings closes some of the vocabulary gap. But it degrades when you're asking about specific names and events, and it introduces plausible-sounding noise that's sometimes worse than no result. It's one layer of the solution, not the whole thing.

The gap between search-as-lookup and retrieval-as-recall is still mostly unaddressed in consumer note-taking. The apps that have tried to close it — Mem, Rewind, a handful of others — tend to address the retrieval problem by throwing AI at document summaries. That helps with ambient recall. It doesn't solve the structured-fact problem, the person-preference problem, the kind of retrieval where you're not browsing but querying.

The note you wrote is accurate. The app stored it exactly as typed. And when you press that search button with your best guess at the words you used, the app does exactly what it was built to do.

That's the lie. What it was built to do is not what you were promised.

---

Asgeir Albretsen is the founder of Harbor.