Here's a small tragedy that happens in every research project. Three months ago you read a paper that's exactly what you need right now. You saved it. And you cannot find it — because you're typing "does testing help you remember things," and the paper is titled "Retrieval practice produces durable long-term retention." Same idea. Zero shared words. Keyword search shrugs.
This is the fundamental limit of matching letters: it can only find what you can already name. But the whole reason you're searching is that you can't quite name it — you remember the gist, not the title. So you scroll. You open six PDFs that aren't it. Eventually you re-find the paper by accident, or re-download it, now sitting in your library twice.
We added search by meaning to your Folio library to close this gap. Instead of matching the words you typed, it matches the idea behind them — so it finds the paper that answers your question even when it shares nothing with how you phrased it. Try the same three queries both ways:
Keyword search
A/B testing improves conversion rates
matches “testing” — wrong field entirely
Software testing methodologies
matches “testing” — irrelevant
Search by meaning
Retrieval practice produces durable long-term retention
never says “testing” — but it’s the answer
The testing effect: a meta-analysis of recall
the canonical result
Same query, two engines. Keyword search matches the letters you typed; search by meaning matches the idea — so it finds the paper that answers you even when it shares not one word with your question.
Notice what the keyword column does: it latches onto a word — "testing," "sleep," "putting off" — and drags in whatever else happens to contain it, from entirely different fields. The meaning column ignores the surface words and goes for the concept. That's the difference between "find the string I typed" and "find what I meant."
How it works, briefly and honestly
When you add a source, Folio computes an embedding of its title and abstract — a numeric fingerprint of what the paper is about, learned by a model that has read across the literature. Your query gets the same treatment. Then it's just geometry: the papers whose meaning sits closest to your query's meaning rise to the top. Nothing about the exact words matters; everything about the concept does.
The useful consequence is that you can search the way you actually think:
- Describe the finding, not the title. "The one about spacing out study sessions" will find the spacing-effect literature even if you never learned the term "distributed practice."
- Search in your own words. You don't have to guess the author's vocabulary. Ask your question; get the answer.
- Surface the adjacent. Because it's matching meaning, it also turns up closely-related work you'd never have keyword-matched — the neighbours of your idea.
Keyword search isn't dead
To be clear, exact matching still has its place — when you know the author, the year, or the precise title, you want the literal thing, and Folio still does that. Meaning search is for the other, more common situation: you remember what a paper was about and nothing else. That's most of the time, and until now it was the case your library handled worst.
The quiet win here isn't just faster lookup. It's that your library stops being a folder you have to remember the contents of, and starts being something you can ask. The papers you've collected already contain the answer to your next question. You just needed a way to reach it by meaning instead of by memory of the exact words.
Search by meaning is live in every Folio library now — it's the same search box, just smarter about what you're really looking for. The more you save, the more it has to connect.