No magic. Just structure.

One concept, the best guide on top.

The atom of Bluelearn is a single concept: "the chain rule," "supply and demand," "what a transformer is." Each one can have several guides, ranked by the community and by expert review. The strongest explanation rises to the top and is what you see first, with the ranked alternatives kept visible right beneath it, not ten competing stubs you have to dig through.

Guides are self-contained. They teach the material rather than linking out to it. Subjects like "calculus" or "economics" aren't folders that own pages; they're tags. A single guide can belong to many subjects at once, because real knowledge doesn't sit in one bucket.

Every concept knows what comes before it.

Guides are connected by prerequisite edges. An edge A → B means "understand A before B." Together, every guide and every edge form one global directed graph, not a pile of separate courses, but a single map of how all knowledge depends on itself.

When you hit a guide you don't understand, you don't hit a wall. You follow an edge backwards to the thing you're missing. That's the difference between a dead end and a detour.

The graph is a true DAG. No cycles. A concept can't be its own prerequisite, directly or down the chain. That guarantee is what makes a learning path always finite and always solvable.

Pick a goal. Get the whole climb.

You don't browse Bluelearn one page at a time. You name a destination ("I want to understand gradient descent") and the system walks the graph backwards from it, gathering every prerequisite you'd need, and orders them into a path.

Nothing here is hand-curated. Paths are derived from the graph, so they're always consistent with it. Three ideas do the work:

The top guide isn't the only way in.

Ranking surfaces the strongest guide for each concept, but people learn differently, and some ideas have genuinely different framings. Two structured layers keep that depth visible rather than buried:

Anyone can write. Experts decide what stays.

Open contribution is only useful if quality holds. Bluelearn runs every guide through a pipeline designed to catch duplication early and put correctness in front of people qualified to judge it, before and after publication.

1. Write against the existing graph

   As you title a new guide, the system searches for concepts that already exist, so you improve an existing guide or add a genuinely different take instead of forking a near-duplicate.

2. Submit with a similarity check

   On submission, a similarity pass flags near-duplicates so a reviewer can merge thin overlaps, while genuinely different takes stay on as ranked alternatives. Depth, not a pile of redundant stubs.

3. Pre-publish: verifiers

   Before a guide goes live, qualified verifiers review it against a published rubric: accuracy, prerequisites, clarity. It publishes when it clears the bar, not on a timer.

4. Post-publish: moderators & votes

   Live guides keep earning their place. Moderators and community votes (each tied to a specific rubric reason, not a thumbs-up) surface what needs work.

5. Disputes

   When experts disagree, a structured dispute resolves it in the open, with the reasoning recorded, so the decision is auditable later.

Open doesn't mean unguarded.

A platform anyone can write to needs clear, fast levers for content that legally or ethically can't stay up. Bluelearn tracks these as content holds, explicit categories like copyright (DMCA), CSAM, and non-consensual imagery, that can take material down quickly and on the record.

Privacy is handled the same way: account erasure works by anonymizing your contributions rather than deleting the shared knowledge they're now part of. The guides stay; your identity doesn't. See the privacy policy and terms for the specifics.