What the research now knows how to do.
The experiments produced more than phenomena. Some structural results have become reusable, tested candidate abilities. Others remain frontiers, and some scientific machinery should remain scientific machinery.
Seven candidate abilities
Candidate means usable and evidence-backed, not frozen or universally applicable. Each ability carries a boundary learned from the research that produced it.
Relate without merging identity.
Keep both endpoints exact and distinct while recording the kind, scope, law, context, evidence, and effective time of their relationship.
Make relation behavior explicit.
Composition, inverse, entailment, integrity, symmetry, and effectivity belong to a declared law profile rather than hidden assumptions.
Know the difference between no and unknown.
Resolve a mapping as matched, not applicable, explicitly negative, conflicting, ambiguous, or unknown instead of flattening all failures.
Inspect the path, not only the endpoint.
Compose witnessed relations in declared order and compare a direct claim with the path that derives it.
Declare what must survive.
Separate successful execution from semantic preservation by recording each required invariant as preserved, lost, unknown, or not applicable.
Use the law that was effective then.
Reconstruct historical meaning without silently substituting today’s relation law for the one governing the original assertion.
Leave an inspectable receipt.
Return the relation, context, target, evidence identity, preservation outcome, and loss without mutating either endpoint or creating institutional authority.
A transformation can execute successfully while a preservation contract fails. “It ran” and “what mattered survived” are different facts.
What comes next
These are research directions, not APIs advertised ahead of evidence.
Semantic law supersession
Experiment 018 falsifies timeline-only retirement: historical validity, current admissibility, and downstream transition action are separate. API extraction remains withheld pending broader replication.
Semantic change impact
Experiment 019 falsifies raw reachability as a sufficient blast-radius rule. Typed causal dependencies can propagate stale, migration, review, cease, history, invalid-new-use, and unknown outcomes while references and same-subject links remain unaffected.
Local-to-global semantic assembly
Experiment 020 assembles Legal, Operations, and Audit without a master schema when required overlap roles are connected and alternate translation laws cohere. Missing overlap, contradictory observations, and off-sample witness-law defects block assembly.
Perspective drift & authority
Experiment 021 separates revision recency from semantic authority. Authority is role-scoped and time-aware; stale projections remain visible without overriding current meaning, while authority conflict/gaps fail closed and contestation triggers review.
Partial-trust / privacy-constrained federation
Experiment 022 projects local views into bounded disclosure packets. Exact disclosure, bounded attestation, deliberate nondisclosure, policy-prohibited disclosure, and genuine unknowns remain distinct; trusted support can sustain a qualified assembly without centralizing every raw field. This is not a cryptographic privacy claim.
Verifiable disclosure & trust revocation
Test what happens when previously accepted bounded evidence later loses trust or authority: historical reliance must remain reconstructable while current admissibility may fail closed.
Agent path coherence
Test whether alternate tool/action paths realize the same intended semantic effect, especially when order changes the result.
Research boundary. Berry phase, Chern topology, non-Abelian holonomy, Yang-monopole calculations, patchwise gauge numerics, and related scientific machinery remain Geometric Transport research capabilities. Their structural lessons can be extracted without relabeling the scientific operators as generic semantics.