[domijin]

Summary

Bootstraps Open Learning Lab as a minimal research system rather than a product scaffold.

Adds

• project charter and north-star objective;
• agent-neutral protocol;
• methodology, architecture, governance, privacy, and glossary docs;
• canonical schemas for goals, hypotheses, experiments, evidence, findings, expert judgment, and decisions;
• append-oriented public ledger with G001, H001-H004, E001, V001-V002, and D001;
• Marin as Study 001 with source policy and decision log;
• learning-session, retrieval-test, and expert-review protocols;
• GitHub issue/PR templates;
• CI integrity validation for JSON, IDs, and key record references.

Research stance

Only H001 is active. H002-H004 are proposals.

The initiating Marin session is recorded as evidence, but no finding claims H001 is validated. Delayed unassisted retrieval/transfer and an eventual comparator remain necessary.

Review questions

1. Is the public/private boundary appropriate?
2. Are the canonical objects minimal enough?
3. Does the ledger preserve process without overfitting to one tutor/model?
4. Are we treating expert judgment as contextual evidence rather than authority?
5. What should be removed before adding any UI/backend?

Related hypotheses: #1 #2 #3 #4

[domijin]

LGTM