🕯 the map admits where it ends
Hands, words, learners
One named walk. Three cards. No hidden answer.
This permanent address carries one three-card walk, notice where observation ends and inference begins, then stop. Leaving with a better unknown is complete.
chosen trail
Hands, words, learners
Question joining the three: What must remain true when a skill moves from a page, through an agent, into different hands and languages?
Every link is optional. Open at most three links per card; stopping earlier is complete.
1 of 3
Mindicraft edge · reviewed 2026-07-24
Which guide would change first if its steps met real hands, materials, and weather?
When hands answer back
Open card 1 of 3: When hands answer back
Known here
- The build labels a guide hand-tested only when its evidence record says hand-tested; every other guide is labelled tradition-asserted.
- At review time, the committed evidence registry contains no entries.
Still unknown
- Which instructions fail first across different materials, climates, bodies, and skill levels.
- How much independent observation is enough to call a practical guide hand-tested.
Evidence that would move it
- Independent field notes that name the setting, materials, exact step, observed outcome, and correction.
- A repeat test showing whether the correction travels to another setting.
Three lenses, at most
- Which step makes the most assumptions about material, climate, or prior skill?
- What observation would show that the wording, rather than the craft, caused a failure?
Paths in: How to use this guide · Keep records and accounts · Teach the young everything here
Sources: Mindicraft — Mindicraft Guide builder · Mindicraft — Mindicraft evidence registry
2 of 3
Mindicraft edge · reviewed 2026-07-24
When one physical skill crosses four languages, what must remain invariant?
Four tongues, one physical skill
Open card 2 of 3: Four tongues, one physical skill
Known here
- The same guide slugs and tree relations are shared across Mindicraft's four language editions.
Still unknown
- Whether every translated warning points to the same physical sign, boundary, and action.
- Which local materials or customs make a literal translation technically misleading.
Evidence that would move it
- Bilingual review tied to a shared object, image, measurement, or observed outcome.
- Hand tests in more than one language and place, with disagreements kept rather than averaged away.
Three lenses, at most
- Which physical referent must stay the same when the words change?
- Where might a locally correct explanation need different examples rather than parallel wording?
Paths in: Tie knots and lashings that hold · Make soap from fat and lye · Plant a village
Sources: Mindicraft — Mindicraft language registry · Mindicraft — Mindicraft technology tree
3 of 3
Mindicraft edge · reviewed 2026-07-24
Does a lesson an agent reconstructs easily predict what a human can learn?
Two windows on learning
Open card 3 of 3: Two windows on learning
Known here
- Mindicraft can return a guide with its prerequisites as prompt-ready text, but its evidence registry records no paired agent and human learning outcomes.
Still unknown
- Whether fluent reconstruction by an agent predicts correct later performance by a person.
- Which parts of teachability are shared across model, learner, body, and environment.
Evidence that would move it
- Paired tests on the same skill that separate restatement, procedural accuracy, delayed recall, and independent performance.
- Results across multiple guides rather than one unusually easy or difficult lesson.
Three lenses, at most
- Which success measure belongs to the words, and which belongs to later action?
- What result would make agent fluency a poor proxy for human learning?
Paths in: How to use this guide · Teach the young everything here · Keep records and accounts
Sources: Mindicraft — Mindicraft Guide builder · Mindicraft — Mindicraft evidence registry
Castle map:- Castle map: two windows Link only — NOASSERTION; no licence grant
- Castle map: teachability validated Link only — NOASSERTION; no licence grant
Three cards is complete. Leaving with an unknown is complete.
The three questions carried through every walk
- What does the source actually show?
- Which part is inference or remains unknown?
- What one result would move the boundary?
Optional return shape: noticed · inferred · still-unknown. Keep it with you or give it to whoever invited the walk; these
pages receive nothing.
A carryable trail. This stable address can be
bookmarked, printed, or shared. Its path contains only a reviewed public trail name—no identity, note, answer,
seed, or result. This page receives no response. An agent can request the same address with
Accept: application/json, or read
its direct JSON file.