Research profile

Verifiable Civilizational Learning

Testing whether verified experience can accumulate outside model weights and survive complete agent and model turnover.

Institutional learning

Testing whether verified experience can accumulate outside model weights and survive complete agent and model turnover.

Verifiable Civilizational Learning treats an institution's verified experience as a possible unit of learning. Persistent identities, bounded actions, execution evidence, review, and versioned precedent form a loop through which experience may become reusable skills, tests, policies, and routing constraints. Its decisive proposed experiment replaces every contributing agent and model lineage, then measures whether successor systems retain useful capability and defined constraints. Comparisons include isolated models, individual memory, and ordinary shared-memory agent systems.

The research question

Can outcome-verified capability transfer across model families and survive total turnover of the agents and models that created it?

Research workstreams

A question. A method. A test.

Workstream 01

Outcome-gated lesson and precedent promotion

Workstream 02

Cross-substrate transfer and full-turnover benchmarks

Workstream 03

Repeated-failure, integrity, constraint-retention, and compute-normalized metrics

Evidence priorities

Make progress measurable.

The initiative is framed around testable questions. Its research direction should be assessed against observed evidence and the limits of each experiment.

Capability retained after all contributing agents and model lineages are replaced
Constraint retention and reduction of repeated failures
Comparison with individual memory and ordinary shared-memory systems

Research goals and proposed mechanisms are not established findings. Product-linked tracks describe the research behind a capability, rather than a claim that its hypotheses have been validated.

Associated company