Find your next learning step
Retiring a topology claim should stop the same evidence search from returning under a new label. A successor can open only after a boundary review shows what is scientifically different and what lineage remains shared. Fresh evidence is then designed outside the retired family’s exposed data and repair choices. The archive stays useful for provenance, training and falsification, but it cannot quietly become confirmation for the successor.
Open the chapter contents · Topology root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the decision, or read from Chapter 1. The chapters remain open; these grouped routes are a navigable index.
Route 1 · Define the decision
Route 2 · Work the ledgers
Route 3 · Transfer the method
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · DEFINE THE DECISION
1. Define the retired claim precisely
Write the proposition, domain, topology, channels, decision rule and authority that ended. Retirement of an active-set unlock claim is not retirement of topology as a method or every descriptive use of the archived graph.
A precise boundary prevents two errors: treating a genuinely new question as forbidden forever, and relabelling the exhausted claim as new while keeping its data and selection path.
CHAPTER 2 OF 12 · DEFINE THE DECISION
2. Test whether the successor is substantively new
Compare endpoint, population, topology, mechanism, channel pool, evidence source, decision consequence and analyst choices. Cosmetic changes, threshold tweaks and one extra channel usually remain in the old family.
A successor may share code or notation while making a different claim. Record shared lineage and explain why old failures do or do not constrain it. Independence is not created by a clean repository.
CHAPTER 3 OF 12 · WORK THE LEDGERS
3. Build a successor-boundary matrix
A team proposes rule S after retiring restart family R5.
| Boundary dimension | Retired R5 | Proposed S | Shared dependence | Review |
|---|---|---|---|---|
| Endpoint | Fast unlock | Stable component discovery | Partial | Potentially new |
| Topology | Adaptive active set | Fixed preregistered graph | Low | New structure |
| Channels | Selected after failures | Prospective fixed pool | None if enforced | Fresh |
| Population | Rounds 1–220 | New operating regime | Some mechanism overlap | Sensitivity needed |
| Decision authority | Automatic unlock | Advisory flag | Different consequence | New claim |
| Analyst team | Development group | Independent audit group | Process separation | Document |
The panel permits a prospective advisory study but does not authorise automatic unlock authority.
CHAPTER 4 OF 12 · WORK THE LEDGERS
4. Work a fresh-evidence allocation
The successor receives 120 new rounds split before outcomes: 40 for method setup, 50 for locked confirmation and 30 for independent stress testing.
| Evidence block | Rounds | Rule visibility | Permitted role | If boundary breaks |
|---|---|---|---|---|
| Setup | 1–40 | Editable | Development | Relock protocol |
| Confirmation | 41–90 | Frozen | Primary evidence | Stop and report |
| Stress test | 91–120 | Hidden from developers | Independent challenge | Review claim |
| R5 archive | Historical | Fully exposed | Falsification and context | Never confirmation |
| Future extension | Unallocated | Undefined | None | New review |
If setup consumes information from the confirmation block, those rounds move to development and are not replaced by a post hoc relabelling.
CHAPTER 5 OF 12 · WORK THE LEDGERS
5. Design evidence outside the old search
Use new outcomes, prospectively fixed channels, hidden stress cases and predictable analysis rules. Separate developers from the final challenge where practical. Record every exposure, override and protocol change.
Historical cases remain valuable for checking that the successor does not repeat known failures. That is a falsification or robustness role, not clean evidence for the advertised success rate.
CHAPTER 6 OF 12 · TRANSFER THE METHOD
6. Govern post-retirement archive use
Permit reproducibility, provenance, training, failure taxonomy, sensitivity analysis and successor falsification. Restrict operational control, selective success demonstrations and any use presented as fresh confirmation.
Attach visible status, retired claim, effective date, permitted roles and successor link to the archive. Log exports when exposure could contaminate a future audit set.
CHAPTER 7 OF 12 · TRANSFER THE METHOD
7. Changed-context transfer: adaptive screening graph
Transfer the method to a screening graph retired after repeated edge repairs. A proposed successor uses a fixed graph for an advisory flag on a new cohort rather than an automatic diagnostic action.
The boundary review accepts the advisory claim but keeps historical patients in the exposed archive. New patients and a hidden stress set supply confirmation. Any future move to automatic action requires another authority review.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Connect family stopping, restart control and provenance
Start with family-wide stopping rules and claim retirement. Use restart-family limits and repair-sequence dependence to recover the exhausted search.
Return to reinstated-rule monitoring and multiple-restart error control when a proposed successor still shares the old operational authority.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Limits and common misconceptions
New data are not automatically fresh if choices were shaped by previewed outcomes or the same adaptive search. A different endpoint is not automatically independent when it is selected because the retired endpoint failed.
Archive use is not contamination in every role. It becomes a problem when exposed evidence is described as independent confirmation or quietly changes a supposedly locked successor. Boundary review is governance, not a proof of universal statistical validity.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Practical evidence and access checks
Archive the retired claim, family ledger, failure reasons, exposed rounds, channel choices, evidence spend, permitted uses, access controls and successor proposals.
For the successor, record boundary matrix, intended authority, preregistered topology, new evidence source, block allocation, stopping rules, independent roles, contamination incidents and the conditions for later deployment or retirement.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Frequently asked questions
Can the successor reuse archived code?
Yes, if shared implementation is documented and fresh evidence tests the new claim; code reuse does not make evidence independent.
Can archived failures be used in stress testing?
Yes. They are valuable adversarial cases, but performance on exposed failures should not be sold as confirmation.
Who decides that a successor is new?
The named review authority should include independent scientific and operational judgement, with conflicts and prior data access recorded.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Sources and next reading
The W3C PROV-O Recommendation provides a formal vocabulary for entities, activities, agents, derivation, generation and invalidation that can represent retired and successor lineage. Ramdas and colleagues explain safe anytime-valid inference with e-processes; validity depends on the stated evidence process, not a renamed analysis. Vovk and Wang’s work on sequential e-value merging illustrates the care required when combining sequential evidence. The boundary matrix and allocation are original teaching examples.
Continue through the Topology root, the Gold Standard of Education and How X Works.
Top · Contents · Previous · How X Works