Dashboard › institutional-transition-lab › Distillation
7ab5d3bc-44f0-4c25-b60d-b23cbefbf7bd["lore_tm_v1_YPlsyJRIFRyLJQX1VoF_S_icJZF9jug-KvwSvy45bHw","lore_tm_v1_i_BsbgaYWwlvNVBinlK_iihk_b37BCsO9KatoxkeTTk","lore_tm_v1_S4YL6YwSjeEKH4SuqV2rd8dGRPhpN3LmLkOGYiLuc6g","lore_tm_v1_mnw5sutVQwfrxhhpr8RemvnmWd76EuusIaYZpgd4deY","lore_tm_v1__vTOq9L9U-3I26a4o7F7ggKkZlNDovmC1V0kXTKepXk","lore_tm_v1_fCnSPWsqe2KcWlPNP5fzACwBUkuyMmi6_VlVuJXV37U","lore_tm_v1_FnaiGzRJWhRHj4OS0OpH5aBgJCyZu8xgZDEK3CVA0Wo","lore_tm_v1_Q5j3q6SyKKomJAA-u8Lhivkw2dtpxaNOhwmrTeGNFMw","lore_tm_v1_aPIOZ5lQ_iJ696uVZzxcK39Wb9-VkLsIuLmC9Oiu9dc"]
/home/byk/Code/institutional-transition-lab/research/oss-governance-coding-protocol-v1.md, /home/byk/Code/institutional-transition-lab/research/oss-governance-coding-protocol-v1.1.md, and /home/byk/Code/institutional-transition-lab/research/oss-governance-coding-protocol-v1.2.md./home/byk/Code/institutional-transition-lab/research/session-handover-2026-08-27.md states institutional-power changes should precede project-performance changes; references zero records in a canonical institutional power ledger; requires grounded evidence for every accepted event, date, actor, and right; reports strict power-edge recall varying from 75% to 91.7%; and excludes power-graph serialization, causal identification, and performance-outcome work from the current scope./home/byk/Code/institutional-transition-lab/research/sol-reasoning-hard-case-protocol.md asks whether gpt-5.6-sol improves institutional-event discovery or strict power-edge performance; excludes adjudication, event-ledger, and canonical-power-graph construction; evaluates lower-budget value for power-change recall; and tracks deviations, decision and power-match unanimity, and latency per document./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v2.md identifies Valkey PR #345 dated 2024-04-30 as a false negative: it created GOVERNANCE.md, established the TSC, and assigned oversight, appointment/removal, voting, delegation, and override rights./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v2.md identifies OpenTofu PR #2830 dated 2025-05-23 as correctly queued with Abstain but not extracted; its merged charter and governance files explicitly allocate TSC oversight, appointment, removal, voting, delegation, licensing-exception, and trademark rights./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v2.md identifies Valkey PR #3543 dated 2026-06-01 as a false positive: it is a 53-file Raft-based cluster-protocol implementation rather than an organizational-rights change./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v2.md says model review preserved additional clear rights-change records despite wrong power-change coding and that the review queue contained four such clear rights-change records; its workflow classifies and codes power edges from an enriched bundle with human involvement./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v1.md targets documents about changes in organizational power while controlling review-queue size; its ontology includes appointment rights, licensing, stewardship, ownership, trademarks, and vetoes; model runs have no tools or memory and must cite every proposed event and power change; and only Terraform PR #33661 among a discussed set changes decision rights rather than merely exercising them./home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v1.md distinguishes technical architecture and usage-right changes from organizational-power changes./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md says no model output was admitted to the canonical event ledger or power graph; reports Luna complete power-edge agreement of 55.0% in one result and 55.0% versus 47.5% in another comparison; and says models were not stable enough to author canonical actor-right-target relationships./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md says substantive product/control events are not institutional power changes; notes exact edge sets differed even where models agreed on event/control-right changes; and specifies canonical fields including relationship, actor, right, target, direction, scope, and evidence references before conversion into dated power-graph diffs./home/byk/Code/institutional-transition-lab/research/luna-terra-hard-case-comparison.md compares model, effort, decision, institutional precision, institutional recall, strict power-edge recall, decision unanimity, power-score unanimity, and mean milliseconds per document./home/byk/Code/institutional-transition-lab/research/luna-terra-hard-case-comparison.md says strict power-edge differences largely reflect serialization choices: Node was represented as either an added merge right, a modified merge right, or added/modified foundation stewardship./home/byk/Code/institutional-transition-lab/research/luna-terra-hard-case-comparison.md reports Terra/xhigh had 100% power-score unanimity and recommends not asking the discovery model to define the final power graph; it references a result with 83.3% strict power-edge recall and 100% evidence grounding./home/byk/Code/institutional-transition-lab/research/luna-event-extraction-development-benchmark.md asks whether events can be translated into auditable power-graph diffs; Run A’s archived raw responses were rescored locally with v3 power-edge labels to compare with Run B; and strict required power-edge recall was 91.7% versus 75.0%./home/byk/Code/institutional-transition-lab/research/luna-event-extraction-development-benchmark.md says excerpts, aliases, and the power-edge ontology were revised after earlier runs; the prompt-injection document did not override the classifier contract; communities’ merge right was encoded as modified in one serialization while foundation stewardship was used in another; and the schema admits multiple defensible serializations of one event./home/byk/Code/institutional-transition-lab/research/luna-event-extraction-development-benchmark.md includes proposing candidate power changes as step 3 and advises against using a single Luna response as the canonical power graph without normalization./home/byk/Code/institutional-transition-lab/research/llm-event-triage-cost-frontier-protocol.md deliberately excludes power-edge recall from acceptance because equivalent changes may serialize into different graph edges; it treats canonical actor, target, and right as separate concerns; and says a 75% edge-recall result does not authorize unaudited power-graph serialization./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-protocol-v1.2.md is titled “Performance-blinded OSS governance coding protocol v1.2,” was frozen on 2026-08-27, and has status pre-adjudication evaluator correction; model-informed development protocol. (meaning August 27, 2026)provisional_exact_agreement.provisional_core_agreement; and 4. reporting full-response agreement separately without requiring it because titles and numeric confidence are not canonical research fields.33060493337, and repeats only the Terra advisory pass over the corrected route to isolate the evaluator change from model sampling variance./home/byk/Code/institutional-transition-lab/schema/governance-coding-v1.schema.json./home/byk/Code/institutional-transition-lab/schema/governance-coding-v1.schema.json uses JSON Schema draft 2020-12, has $id https://github.com/BYK/institutional-transition-lab/schema/governance-coding-v1.schema.json, title Performance-blinded governance coding v1, root type object, and additionalProperties: false.record_class, event_kind, title, announced_on, effective_on, affected_scope, body_patch_relation, power_changes, confidence, and ambiguity.record_class values are effective_institutional_change, announced_institutional_change, proposal_only, control_event, no_event, and abstain.event_kind values are leadership, board_or_steering, control_rights, reorganization, foundation_transfer, license, fork, reunification, strategy, product, and external.body_patch_relation values are consistent, patch_supersedes_body, body_only, patch_only, conflict, not_applicable, and unclear.power_changes item forbids additional properties and requires actor, right_kind, target, direction, change_status, scope, and evidence_refs.right_kind values are appoint, remove, elect, vote, delegate, override, approve, merge, write, release, veto, own, license, steward, fund, set_budget, set_strategy, set_policy, set_membership, set_terms, and inform.direction values are added, removed, and modified; allowed change_status values are effective, announced, proposed, rejected, and unclear.evidence_refs must contain at least one object with exactly the required non-empty string fields source_id and quote; confidence must be numeric from 0 through 1 inclusive..lore.md, README.md, tests, implementation, schemas, research files, and frozen review-case artifacts./home/byk/Code/institutional-transition-lab/.lore.md defines a fail-closed, source-first governance-adjudication rule: never infer organizational authority or effectiveness beyond bounded source evidence; software ownership, implementation responsibility, technical control, and compatibility language are insufficient to establish organizational power or decision rights..lore.md requires proposals and announcements to remain distinct from implemented changes; unmerged, rejected, or open proposals must not receive effective dates; implementation evidence must not establish an underlying third-party policy’s effective date; every institutional claim and graph edge must use exact verbatim quotes with source IDs; conclusions must remain within directly evidenced scope; unsupported actors, rights, transfers, and enforcement claims must be removed; truncation, unavailable patches, unstable history, and other evidence limitations must be reported; and adjudicators must abstain when bounded evidence is insufficient./home/byk/Code/institutional-transition-lab/README.md says 35 records are routed to source adjudication and only five have provisional core agreement./home/byk/Code/institutional-transition-lab/schema/governance-adjudication-v1.schema.json requires adjudication record fields including record_id, source_url, decision, accepted_from, reviewer_ids, claim_evidence_refs, final_coding, and graph_edge_revision; decision allows accept, revise, reject, and abstain; accepted_from allows luna_a, luna_b, terra_advisory, or null./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py validates that accept requires accepted_from to name a valid frozen response and requires final accepted coding to equal that response; only accept may set accepted_from.governance_adjudication.py requires resolved accept, revise, or reject decisions to carry graph_edge_revision; requires abstain to use null graph_edge_revision; requires an accepted revision’s basis to equal accepted_from; and checks that graph_edge_revision.added and .removed are derived.governance_adjudication.py emits graph_edge_revisions in its summary as objects containing record_id plus each record’s revision fields./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md has status frozen development result; source adjudication required, says institutional candidates require source adjudication, and notes V1.2 corrected the evaluator before adjudication while reusing frozen materials./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-protocol-v1.md requires reporting zero canonical-ledger records pending human/source adjudication and later actor work./home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1/v1.2-coding-package.json contains the title Valkey governance and TSC decision rules revised./home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1/documents.json contains frozen source bodies and patches, including OpenTofu TSC/RFC material, an OpenTofu CONTRIBUTING patch that changes contribution assignment prerequisites, and a Valkey Samples Repository RFC describing its proposed design, governance model, contribution framework, intended audiences, content, review process, and structure./home/byk/Code/institutional-transition-lab/tests/test_governance_adjudication.py fixture _adjudication() uses schema_version: 1, adjudication_id: "oss-governance-adjudication-v1", design_status: "performance_blinded_llm_assisted_source_adjudication", transition_dates_excluded: True, and outcome_data_used: False._adjudication() fixture pins workflow_run_id: 33063470968, artifact_id: 9642983578, plus artifact_sha256, documents_sha256, documents_file_sha256, coding_package_file_sha256, protocol_sha256, coding_schema_sha256, and adjudication_schema_sha256.opencode-gpt-5.6-sol-primary, reviewer type llm, model openai/gpt-5.6-sol, role source_adjudicator, with assistance description Primary source review and synthesis.example-1 record accepts luna_a, uses reviewer opencode-gpt-5.6-sol-primary, gives rationale The source contains no institutional transition., has no evidence limitation or unresolved reason, uses empty claim_evidence_refs, and sets graph_edge_revision to basis luna_a with empty added and removed arrays._validate() in /home/byk/Code/institutional-transition-lab/tests/test_governance_adjudication.py computes a canonical documents hash with canonical_sha256(documents), writes it into the coding-package summary, adjudication inputs, and verified pins, then calls validate_adjudication(adjudication, documents, package, verified_pins=pins).test_one_abstention_keeps_canonicalization_gate_closed() changes the fixture to decision: "abstain", accepted_from: None, unresolved reason Required source evidence is incomplete., final_coding: None, and graph_edge_revision: None; expected outcome is no validation errors, record_count == 1, unresolved_record_ids == ["example-1"], and canonicalization_gate == "closed".test_record_ids_must_match_frozen_documents_exactly_once() checks both duplicate and missing records; expected validation includes record IDs must match frozen documents exactly once and keeps canonicalization_gate closed.test_accept_must_equal_the_named_frozen_response() mutates accepted coding confidence to 0.5; expected validation includes accepted coding does not equal luna_a and keeps canonicalization_gate closed.test_resolved_decision_cannot_leave_final_coding_abstained() uses decision: "revise" with accepted_from: None, a final record_class of abstain, ambiguity The available source cannot resolve the event., and graph_edge_revision: {"basis": None, "added": [], "removed": []}; expected validation includes resolved decision cannot use abstain final class and keeps canonicalization_gate closed.proposal_only, control_event, and no_event across /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py, /home/byk/Code/institutional-transition-lab/src/institution_lab/llm_event_probe.py, and /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py./home/byk/Code/institutional-transition-lab/src/institution_lab/llm_event_probe.py defines DECISIONS = {"institutional_event", "control_event", "no_event", "abstain"}; treats institutional_event and control_event as events; requires control_event to use product or external event_kind and to contain no invented power change; and requires no_event to have null event fields and empty evidence./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py recognizes proposal_only; date/status validation treats announced_institutional_change and proposal_only specially; and maps proposal_only to allowed change statuses proposed, rejected, and unclear./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py defines RECORD_CLASSES as effective_institutional_change, announced_institutional_change, proposal_only, control_event, no_event, and abstain; EVENT_KINDS and RIGHT_KINDS mirror the JSON Schema enumerations; DIRECTIONS = {"added", "removed", "modified"}; and CHANGE_STATUSES = {"effective", "announced", "proposed", "rejected", "unclear"}.governance_coding.py defines response keys record_class, event_kind, title, announced_on, effective_on, affected_scope, body_patch_relation, power_changes, confidence, and ambiguity; power-edge keys actor, right_kind, target, direction, change_status, scope, and evidence_refs; and evidence-reference keys source_id and quote.governance_coding.py defines MODEL_INPUT_FIELDS in order as: 1. entity_id, 2. publisher, 3. published_on, 4. source_type, 5. sources, 6. evidence_bounds; CODING_DOCUMENT_KEYS adds record_id and source_url.prepare_coding_bundle() in governance_coding.py joins source bodies to enriched records without model-side triage metadata, requires list-valued records and documents, requires each enriched record to have a matching source document and non-empty text, and creates a base source with source_id: "body", kind: "github_body", filename: None, and the body text.pull_request_state_and_patch, prepare_coding_bundle() serializes metadata fields state, draft, merged, merged_at, merge_commit_sha, base_sha, head_sha, and changed_files into source pr_metadata; adds available relevant-file patches as patch:{index} sources of kind github_pr_patch; and counts unavailable patches.prepare_coding_bundle() records evidence bounds source_text_truncated, files_listing_complete, patch_selection_truncated, and patch_unavailable_count; non-PR records use files_listing_complete: None, patch_selection_truncated: False, and patch_unavailable_count: 0.prepare_coding_bundle() outputs schema_version: 1, coding_id: "oss-governance-enriched-coding-v1", the supplied coding_schema_sha256, design_status: "performance_blinded_model_informed_development", transition_dates_excluded: True, outcome_data_used: False, the ordered model_input_fields, and prepared documents.parse_response() strips optional fenced JSON, rejects unterminated code fences, parses JSON, and requires the result to be an object.validate_response() requires the response key set to equal RESPONSE_KEYS; validates allowed record classes, event kinds, body/patch relations, ISO dates, and non-empty-or-null title, affected_scope, and ambiguity; and requires finite numeric confidence from 0 through 1 while rejecting booleans.validate_response() requires power_changes to be an array and each edge to contain exactly POWER_KEYS; validates non-empty actor, target, and scope; validates allowed right, direction, and status values; requires non-empty evidence_refs; and considers a reference grounded only when its source_id resolves to source text and its non-empty exact quote occurs within that text.validate_response() defines institutional records as effective_institutional_change, announced_institutional_change, or proposal_only; these require at least one power change, while every non-institutional class must have no power changes; abstain additionally requires non-null ambiguity._validate_run() requires a non-empty coding-document list, rejects malformed or duplicate raw responses, requires raw response IDs to match coding-document IDs exactly, and validates coding_id, coding-schema hash, run_id, source_url, the exact MODEL_INPUT_FIELDS allowlist, model, reasoning effort, and nonnegative integer elapsed_ms._validate_run() parses each response_text, records JSON or schema errors, and emits per-record results containing at least parsed response, schema_valid, validation_errors, and model.