Relational Atlas views improve explanation and transfer.Unsupported
Required next test: A preregistered, content-matched comparison with strong flat and learner-built views.
Companion version 0.4 · Snapshot 31 August 2026 · Book draft 0.7
Worlds Outside the Mind
The living product, evidence, governance, and correction companion to Axel Pond's short research book. Transient facts live here so the book's intellectual argument can remain stable and product-independent.
What belongs here
This page records current feature scope, implementation status, educational hypotheses, governance gaps, negative results, and corrections. The book contains only the stable theoretical argument and an optional afterword explaining why Noosaga and Nousfarer are relevant design experiments.
Noosaga's Atlas is a comparatively direct application of relational organization, provenance, plural projection, and correction. Nousfareris the further hypothesis that place, movement, play, and personal discovery can add value beyond a strong nonspatial interface. Book version 0.7 frames both around progressive disclosure: orientation first, with deeper relations, sources, alternatives, and provenance available when the learner's task calls for them.
Current product snapshot
- Noosaga currently describes an Atlas of 1,800+ fields across 200+ disciplines and 19 top-level categories.
- Atlas workflows support subfield maps, framework timelines, articles, curricula, prerequisite graphs, vocabulary timelines, relations, and on-demand concept material.
- Nousfarer currently supports bridge → category galaxy → discipline system → subfield planet → surface → framework location → concept room → article and quiz progression.
- The ship includes a private museum and a discovery-and-placement system; the Lion-Man is the starter sculpture used in onboarding.
- Generated or mechanically valid learner content is not thereby manually approved. Review state belongs to the exact persisted content version.
The journey is a hypothesis, not a result. Remembering a place does not establish understanding of the field it represents.
Governance and provenance
These labels report actual intermediate state. A specified commitment is not presented as an implemented virtue.
| Commitment | Status | Missing work |
|---|---|---|
| Source requirements for factual framework claims | Partially implemented | One public, claim-type-specific policy and audit view. |
| Visible generated, automated-review, and manual-review states | Partially implemented | Consistent labels on every learner-facing artifact. |
| Correction tools and preserved revision history | Partially implemented | A unified public history, response targets, and appeal path. |
| Contested classifications and multiple parentage | Specified but not implemented | First-class plural projections with visible competing rationales. |
| Regional, linguistic, and disciplinary gap reporting | Specified but not implemented | Reproducible metrics and a public coverage dashboard. |
| Reproducible dated dataset snapshots | Specified but not implemented | An export contract, release cadence, and archival location. |
Educational claims
Product presence, engagement, and automated-pipeline status are not learning evidence.
Nousfarer's spatial hierarchy improves orientation and delayed conceptual location.Unsupported
Required next test: Compare it with a strong 2D map while measuring novelty, time, presence, and cognitive load.
Museum placement and revisiting improve interpretation and retention.Unsupported
Required next test: Compare it with unstructured favorites and non-collection review tools.
Generated explanations are educationally sound when automated checks pass.Unsupported
Required next test: This inference is rejected; quality requires exact-content manual review tied to a content hash.
Negative result: Agora
An earlier experiment supplied several language-model roles with Atlas context. Internal comparisons did not show an advantage over strong direct-model baselines sufficient to justify the added latency, cost, and complexity, so the feature was deprecated.
This is a documented product decision, not formal empirical evidence. A publishable result requires a complete archival report of tasks, models, prompts, conditions, sample size, criteria, outputs, uncertainty, exclusions, latency, cost, and code version.
Claim ledger and corrections
The versioned source release includes a claim ledger separating established ingredients, interpretive syntheses, novel hypotheses, methodological rules, broader implications, product facts, and design aspirations. Its open objections include distinctiveness, measurement circularity, content matching, causal scale, power, cross-cultural authority, and normativity.
Corrections and source challenges: axel@noosaga.com