Repository navigation
docs(atlas): add scarcity-aware rubric and seven-variant blog - #19
Merged
Merged
Conversation
Interpretation-layer and documentation only. No code, tests, configs, protocol, registration, universe, identity map, locks, packs or the original run record changed. No panel selected, no panel approved, no human approval claimed or implied, no classification, no RescueScreen, compound or treatment claim. The audited 2026-08-06 machine result is preserved byte-identical: run record 5f5b0918a24fcaa737877a15e393773f20c584c531517133e9c8b7e7574cfffd still reads INFEASIBLE_PANEL under atlas-phase2-panel-selection-protocol 1.0.4 (24/24 attempts INFEASIBLE_COMPLETE at sizes 7/6/5 across L0-R7, max 714 nodes against a 5,000,000 budget, zero relaxation steps, zero members, 35 complete dispositions). Protocol v1.0.4, its registration, universe lock v4 and identity-map lock v4 are untouched. ADR-0019 adopts versioned interpretation rubrics with named context profiles. Precedent one is ADR-0013, which split a coarse FAIL into integrity, sufficiency, performance, policy and authorization axes while preserving R2 byte-identical. Precedent two is the Atlas panel: v1.0.4 conflated "is a strictly independent panel feasible" with "is any rare-disease panel scientifically useful". The response is additive, symmetric and post-hoc, and the governing rule is that a rubric is never tuned to make one observed result look better. ADR-0019 supersedes the interpretation policy only, never ADR-0017's preserved formal result or its two-product separation. docs/project/specs/atlas-panel-rubric-v2.yaml defines atlas-panel-rubric 2.0.0, profile rare-disease-specialist-literature-v1, status POST_HOC_INTERPRETATION_ONLY: fifteen independent axes (AX-1..AX-15), an eight-value closed outcome vocabulary (INDEPENDENT_VALIDATION_PANEL, EXPERT_ADJUDICATED_RARE_DISEASE_PANEL, TECHNICAL_COVERAGE_PANEL, EXPERT_ADJUDICATION_REQUIRED, PANEL_NOT_JUSTIFIED, MORE_EVIDENCE_REQUIRED, INFEASIBLE_UNDER_RUBRIC, UNDETERMINED_SEARCH_INCOMPLETE), nine non-waivable dimensions (NW-1..NW-9), four contextually adjudicable dimensions (CA-1..CA-4 = P2, D3, D2, panel balance/size), twelve required adjudication-record fields, the ten-step evolution workflow RE-1..RE-10, major/minor/patch versioning, seven mandatory output carry fields and a published transition matrix. Scarcity lowers claim ceilings and never strengthens evidence; no expert can convert dependent evidence into independent replication or waive a non-waivable dimension. data/atlas/tsc2_phase2_panel_rubric_v2_readjudication_2026-08-15.json is a candidate-free post-hoc record with a self-excluding canonical content hash bf043a2cb1f195bcf82c412e008862e46565b64bf227db7c679bac9d099d88dc computed over sorted-key, indent-2, LF-terminated JSON with the hash field removed. It binds the source run by path and committed-blob hash and copies only aggregate facts (35 records; 25 resolved / 10 unresolved; 10 access-blocked X5; 1 anchor X3; 14 eligible but unselected; 24 attempts; max 714 nodes / 5,000,000 budget; zero selected members; 37 observations pooled as LG:UNKNOWN-POOL with zero established groups). It preserves machine_outcome INFEASIBLE_PANEL and ADDS contextual_outcome EXPERT_ADJUDICATION_REQUIRED with panel_selected false, panel_approved false, expert_adjudication PENDING, independent_validation NOT_ESTABLISHED, claim ceiling NONE_NO_PANEL_EXISTS and prospective status POST_HOC. Assay/model substrate characterizations and the P2/D3 counterfactual are explicitly labelled post-hoc rather than as fields of the frozen run, and the counterfactual is DIAGNOSTIC_POST_HOC_NOT_A_SELECTION with empty membership. docs/blog/2026-08-15-seven-variants-human-review-boundary.md explains the Mechanism Atlas and Gates 1-8, names the seven TSC2 variants (Arg611Gln, Ile723Val, Arg1743Gly, Thr509Pro, Leu916Pro, Asp1656Tyr, Arg308Trp) and their eight claim-level Gate 8 decisions with exact verified spans, notes the shared Thr509Pro/Arg1743Gly joint-statement span and the Asp1656Tyr minigene/splicing context, separates the anchor plus six-variant repeatability cohort from the 35-record formal panel attempt, states the exact no-panel cause, explains why specialist-lab dominance is expected rather than bad evidence, and reports both outcomes side by side with hashes and a research-only boundary. All packages passed Gates 1-7, Gate 8 is pending, accepted claims are zero, and none of it is a classification. METHOD.md gains a versioned rubric-evolution section; PROGRAM.md reports the machine and contextual outcomes side by side; README.md shows machine result INFEASIBLE_PANEL under v1.0.4, post-hoc rubric-v2 interpretation EXPERT_ADJUDICATION_REQUIRED and no panel selected or approved, and links the new post. TODOS.yaml completes the rubric-v2 design, readjudication and blog tasks, adds the pending named expert panel-composition adjudication kept distinct from human claim-span review, and leaves independent-validation source work and every implementation or screening task uncompleted. Files: docs/project/specs/atlas-panel-rubric-v2.yaml (new), data/atlas/tsc2_phase2_panel_rubric_v2_readjudication_2026-08-15.json (new), docs/blog/2026-08-15-seven-variants-human-review-boundary.md (new), docs/DECISIONS.md, METHOD.md, docs/PROGRAM.md, README.md, docs/project/TODOS.yaml. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 90703341-0147-4b5c-9939-42be522500b1
Correct the rubric enum and coverage wording, publish the exact post-hoc constraint matrix, and narrow the public P2/D3 explanation to the R6/R7 modified constraint set. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 7c146921-f3dd-4a1e-8cf0-8f574de49204
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
INFEASIBLE_PANELmachine result under protocol v1.0.4POST_HOCcontextual layerEXPERT_ADJUDICATION_REQUIREDas a pending state: no panel selected or approvedCritical boundaries
NOT_ESTABLISHEDValidation
5fad899e1ab915d9b3564d871e0ed37d355492402b08153f375e75a281287d4432f085ccf7ee9236a5fda48e7877bd282e137cc679faf3492f70f123f94ab6965f5b0918a24fcaa737877a15e393773f20c584c531517133e9c8b7e7574cfffd