Type a claim. Keep asking why until it stops.
Unfold turns any claim into a plain explanation — then makes every part of that explanation clickable. Open a phrase and it unfolds into its own explanation, one level deeper. Keep going until you hit bedrock, or until we're honest enough to say we don't know.
No invented citations. Every clickable phrase below is re-found, word for word, in the text sitting right above it — never paraphrased, never assumed.
- Verbatim-matched spans
- No invented citations
- Depth-tinted warmth
- Breadcrumbed, always reversible
- Cost-guarded past depth 8
Four small guarantees that make recursive explanation trustworthy
Going deeper is easy. Going deeper without making things up is the actual product.
Verified, not vibes
When the model claims a phrase is clickable, we don't take its word for it. We locate that exact phrase, character for character, inside the explanation it just wrote. If the phrase isn't actually there, it isn't clickable — nothing unfolds on a hallucinated hook.
"...because a single ovary ripens into one fruit..."
"...because the ovary situation explains it..." — not found, not clickable
Depth has a temperature
The first layer you open reads cool — close to the surface's own tone. Go one layer deeper and the card warms slightly. Deeper still, warmer again. You can feel how far you've travelled from the original claim without reading a single label.
Always know where you are
A breadcrumb sits above every open explanation, one crumb per layer. Tap Surface at any point and everything folds back up to where you started — no lost place, no fifteen open panels.
Surface › a single ovary › the flower's female structure
Digging costs something
Recursive explanation isn't a lookup — every new layer is a fresh pass of reasoning. Past eight layers deep, we pause and ask whether you want to keep going, and say plainly why we're asking.
Click a phrase. Watch it go one layer deeper.
A real claim, worked exactly the way Unfold works it. Nothing here is scripted to look smarter than the product actually is.
Bananas are berries, botanically — strawberries aren't.
In botany, a berry is a specific technical category, not a kitchen one — it means a fruit that develops from and ripens with a fleshy wall all the way through, however it tastes. Bananas qualify: the whole fruit grows from a single ovary. Strawberries don't. The red part everyone eats is — flesh built by the flower's base rather than its ovary. The real fruits are the tiny specks on the outside, each a separate one-seeded fruit. So "berry" the biology term and "berry" the grocery-aisle name are simply tracking different things, and were never trying to settle botany in the first place.
An ovary, here, is the part of a flower that holds the ovules — the structures that become seeds once fertilized. "Single" matters botanically: it means one ovary produced the whole fruit, unlike raspberries, which form from many separate ovaries fused onto one flower. Bananas grow from one ovary per flower, and that ovary's wall becomes the fruit's flesh, including the tiny dark specks inside — undeveloped seeds, since most cultivated bananas are seedless by breeding, not by nature. The defining part is simply and how many of them a fruit came from.
This is the pistil — the collective term for a flower's ovary, style, and stigma, the parts that receive pollen and eventually form fruit and seed. It's one of two basic reproductive roles a flower plays, alongside the pollen-producing stamens, and a single flower can hold both roles at once — true of banana and strawberry blossoms alike.
🪨 Bedrock
A defined anatomical term, not a claim under debate — there's no deeper "why" left to unfold, only more anatomy to name.
The receptacle is the tip of the flower stalk that everything else — petals, stamens, ovaries — attaches to. In most fruit it stays small once the flower fades. In a strawberry it does the opposite: it swells into the soft red flesh you eat, while the true fruits, the achenes, stay tiny and dry on its surface. So the part everyone calls "the strawberry" isn't fruit tissue at all — it's fruit grew on top of.
"Accessory fruit" is the standing botanical term for exactly this pattern — edible tissue that ripens from something other than the ovary. Apples and pears work the same way, swelling from the flower's base rather than its ovary wall. It's a recognized, named category, not a one-off exception invented to explain strawberries.
🪨 Bedrock
Settled, citable terminology — the chain ends in a name, not a mystery.
Grocery categories sort by handling, shelf life, and how a shopper searches — not by botanical family. That's why tomatoes sit with vegetables, rhubarb sometimes sits with fruit, and "berry" on a label means small, round, and more than anything from a plant taxonomy textbook. Nobody making these labels consulted a botanist, and nobody was supposed to.
Pint-based selling for small soft fruit predates modern grocery categorization by a long way, and its exact origin — which market, which country, which crop started it — isn't something we can trace to one reliable source. It plausibly grew out of hand-picking and old market-cart measures, but we don't have a citation solid enough to call that settled.
🌫️ Fog
A genuinely open question. We'd rather say so than invent a tidy origin story.
Real sessions go as deep as a claim actually warrants — this demo stops at two layers so the mechanic stays visible. Past eight layers, Unfold pauses and asks whether you want to keep digging, because another honest layer is another real pass of reasoning, not a cached lookup.
Built for anyone who's ever muttered "wait, why though"
No user counts to quote yet — just the kinds of questions we built this to survive.
- The debate-prep friend Wants the steelman, not just the headline.
- The 2am research spiral One tab, then nine tabs, then bedrock, eventually.
- The fact-checking editor Needs the receipt, not the vibe.
- The kid who won't stop asking why The honest, patient version of that game.
- The policy-memo skeptic Reads the footnote before the headline.
- The teaching assistant Explains it once, then again, one level down.
What "verified" actually means here
Every explanation Unfold writes comes back as structured data, not free text we hope stays honest: a paragraph, plus a list of phrases the model wants to make clickable. Before any of those phrases reach you, we run one check — does this exact string exist, character for character, inside the paragraph it's attached to? If yes, it becomes a real clickable hook. If no — if the model paraphrased, summarized, or invented a span that isn't actually there — it's silently dropped.
That's a narrow guarantee on purpose. It doesn't promise the underlying explanation is complete, or that a deeper layer won't eventually run into a real disagreement among experts, or a question nobody's settled. It promises something smaller and checkable: what you can click is what's actually there.
The same discipline governs how far you can go. There's no hard ceiling — claims genuinely do have layers — but past eight levels deep, we stop and ask whether you want to keep digging, and say plainly that another layer means another real pass of reasoning, not a cached lookup.
Some claims stop at one honest paragraph. Some go eight layers down. Both are fine.
Unfold doesn't promise instant expertise — it promises to keep being honest about where the explaining actually stops.