Exploration

What it is. The world is a knowledge-locked space: nothing bars you but understanding. You wander, notice, connect clues, and the “unlock” is a realisation in your head, not an item in a bag. Extension beyond FUN.md’s 21 genres — discovery / immersive-sim-lite, where curiosity is the whole engine.

Player fantasy. “I figured it out.” The rare, clean high of deduction — the world trusted you to reason, and you did. Progress you can’t buy, only earn by paying attention.

Pillars

Knowledge is the only progression. No stat gates, no keys — what you know is what lets you proceed. The player, not the avatar, levels up. A gate opens the moment you understand it (Outer Wilds); it never opens by grinding.
Curiosity is the pull. Every vista, anomaly, and half-answer plants a question. The design’s job is a horizon of “what’s that?” that keeps refilling.
The world is internally honest. It rewards reasoning only if it obeys its own rules perfectly. What you observe must always mean what it seems to — deduction dies on a single lie (Obra Dinn’s fates are ground truth).

The loop stack

Scale	The beat
Moment	Notice → question → probe; a single observation logged.
Encounter	A site/puzzle: gather clues, form a hypothesis, test it.
Session	A thread pulled to a realisation that reframes the map.
Meta	The web of understanding; the endgame that only knowledge unlocks.

Essential systems

System	Why it’s load-bearing
`system-emergent-systems`	A world that remembers and interlocks turns wandering into a personal investigation.
`world-narrative-delivery`	Story lives in the environment/artefacts/systems — embedded, not narrated at you.
`system-onboarding`	Teaching how to notice without hand-holding is the hardest first-ten-minutes problem.
`pattern-emergence`	Interlocking rules let players deduce consequences you didn’t script.
`system-save-and-checkpoint`	A knowledge journal / clue log is the save state that respects the player’s reasoning.

Content & difficulty model

Gate on understanding, not stats. For each gate, name the piece of knowledge that opens it and prove the world teaches it somewhere reachable. This is a connectivity problem: is every required clue reachable before the gate that needs it? (Borrow the roguelike connectivity sweep — see Verify.)
A clue log is the progression track. Surface what the player has deduced (Obra Dinn’s confirmed-in-threes). The journal is player-critical knowledge, exposed — the same discipline as showing “nights left” in a farming sim.
Difficulty is inference distance, not enemy stats — how many clues must chain, how obliquely. Ramp by lengthening the chain, never by hiding a required clue.
Non-linearity is the point. Multiple threads open at once; author a web, not a corridor, and prove several valid orders of discovery reach the end.

Signature-mechanic seeds

Outer Wilds but the loop is a single tide, not a supernova — the world resets on a rhythm you must learn (mechanic-swap; composes genre-rhythm for the clock).
Obra Dinn but the deductions are social — who lied, not who died (theme; composes genre-narrative-decisions for the trust meters).
Exploration but every ability you’d expect to find is instead something you realise you could always do — knowledge as the “power-up” (mechanic-swap; composes genre-metroidvania).
Exploration but co-op: two investigators with different senses must share what only they can perceive (structure; composes genre-coop-chaos).
Exploration but the map is procedural yet its rules are fixed — you learn a grammar, not a layout (structure; composes system-procgen-design).

Common pitfalls

A single lie. If one observed thing doesn’t mean what it shows, deduction becomes guessing and trust collapses. The world must be honest to the frame (the Obra Dinn / tactics perfect-information discipline).
Fake curiosity (fetch quests). Markers and checklists replace wonder with chores. Let the question pull, not a waypoint.
Unreachable required knowledge. A gate whose clue spawns behind that same gate is a softlock. Run the connectivity sweep on knowledge, not just geometry.
Over-narration. Cutscenes that hand you the answer steal the deduction. Deliver through the world; let the player assemble it (see world-narrative-delivery).

Anchors

anchor-outer-wilds — knowledge as the only progression; a curiosity-gated open world that never gates on items.
anchor-return-of-the-obra-dinn — deduction as the core loop; the game trusts your reasoning and confirms it honestly.

Verify — extension note

Exploration is an extension; with no dedicated FUN.md section it composes its parents’ verify patterns:

Reachability/connectivity of required knowledge → FUN.md §10 — Roguelike: assert every required clue/gate is reachable across seeds/orders BEFORE tuning difficulty. Full-knowledge bots prove a line exists, not player experience — exactly the right claim here. This is the primary proof.
Content honesty/lint → FUN.md §21 — Narrative decisions: clues are data — lint for unique ids, every gate’s needs-flag settable, no dangling/contradictory clues, no unreachable content.
Perfect-information honesty → FUN.md §12 — Tactics: what’s shown resolves exactly as shown; deduction demands it.
Determinism + pure-data state (laws 6–7) make the clue web replayable and clone-and-score deducible.

Composes with

system-emergent-systems — interlocking, remembering systems are what make a world worth investigating.
world-narrative-delivery — the delivery half; exploration is its most demanding client.
genre-metroidvania — swap its ability-gates for knowledge-gates and you’re here.