Dashboard › institutional-transition-lab › Distillation
f4aca37a-8078-4d8b-b5ea-1183987cc5d6["lore_tm_v1_0dPZC8FNkRXRXP26YINe9Vn-kY2RDEFIqEMwk-JWw9k","lore_tm_v1_pGl3i0zONWLVhJhXJsO-mW3NW0ale5mePhsSQ1phmIM","lore_tm_v1_u9RoPbPukSxNi-emVTbbQJJv9H9oSWXYbbqtD0HGXKU","lore_tm_v1_OYH2j4tS_iBgjKIr0sJ2CpKPuFhZ-2G7DqP2FElBOj0","lore_tm_v1_W9v_eQSxy6i7Iggukyn67DfBCeejwi4zh_2Ut5br4r4","lore_tm_v1_d11J7jCjsmdXMIkQktOzaO8_ieYby-x1qyYXwHhW2X8","lore_tm_v1_D9KQj-Xf9pwpR7AYM2XpwHDJccDsdD10DywiNwzEE1Q","lore_tm_v1_m8v5yNi4n5i0VHonM5vkdnCQcKrsLPvX4uswKHvJiQ8","lore_tm_v1_RBroYpGHb2jbUyLCSKfxs7di_p5SVELTB1lHvAQ7r80","lore_tm_v1_aHM_pNqq5mEsM3Q3h-OcwUJzbyxV2aYZ_ATNhDwdINg","lore_tm_v1_katiBCnHu-3efwCE0_Wzppr9aYJHtEyqxJiBdTa53sE","lore_tm_v1_gnPsSnwtOlKmCKA63S_CLwLisvmVrbE_c9O01Vhncuc","lore_tm_v1_TuNnFuMZdkkFp3KAMBbolD-BGS_43LnTedyNaB1oAJU"]
research/oss-governance-coding-protocol-v1.1.md is titled Performance-blinded OSS governance coding protocol v1.1, frozen 2026-08-27, with status pre-interpretation execution correction; model-informed development protocol. (meaning August 27, 2026)33056429212 coded 40 records but overlapped the intended pre-protocol envelope on only 39: Valkey #1356 entered and OpenTofu #727 left. The mandatory artifact audit detected this before model classifications were interpreted or adjudicated, so the run is retained only as a debugging and stability artifact rather than the V1 research result.33053908119, artifact 9639245355, at cases/review/oss-governance-enriched-envelope-v1.json.research/oss-governance-coding-protocol-v1.md is titled Performance-blinded OSS governance coding protocol v1, frozen 2026-08-27, with status pre-run, model-informed development protocol; not holdout evidence. (meaning August 27, 2026)schema/governance-coding-v1.schema.json; its SHA-256 is embedded in every coding bundle and response, and the runner refuses execution when checked-in schema and bundle hashes differ.gpt-5.6-luna sessions at medium reasoning independently code all 40 records using the same frozen schema and prompt but sharing neither session state nor model output.0.8; 4. truncated source/patch evidence or an unavailable patch.gpt-5.6-terra advisory coding at xhigh reasoning. Terra is additional evidence rather than an automatic tiebreaker because the prior model comparison did not establish superior Terra accuracy.provisional_exact_agreement. Every output sets canonical_ledger_eligible to false, the run summary must report zero canonical-ledger records, and human/source adjudication plus later actor normalization remain separate stages.research/oss-governance-coding-results-v1.2.md line 136 specifies that the next artifact is a source-level adjudication table covering all 40 records..gitignore ignores .DS_Store, .env, .venv/, __pycache__/, *.py[cod], .pytest_cache/, .ruff_cache/, node_modules/, docs/.observable/, data/raw/**, data/derived/**, *.duckdb, and *.db; it explicitly retains data/raw/.gitkeep and data/derived/.gitkeep.data/raw/ and data/derived/ each contain only .gitkeep.docs/terraform-opentofu.html is titled Terraform and OpenTofu — transition audit; it frames the development case around whether independently detected performance transitions align with changes in control, governance, or leadership and reports no robust transitions.docs/terraform-opentofu.html reads transition_robustness.csv when available, displays candidate_transition_count as Raw candidates and robust_transition_count as Robust transitions, and states that retention requires recurrence within one month across the preregistered robustness settings.institutional_events.csv, assigning event_on from announced_on or otherwise effective_on; its table columns are Date, Entity, Role, Event, Power interpretation, and Source.docs/index.html explains that performance phases are detected without event dates, events are collected across the full period without phase labels, and the two are overlaid only afterward. Its synthetic timeline labels three phases: growth, stagnation, and decline.docs/index.html lists Non-transition periods and placebo events—not just the successful story as item 6 and says a component should be retained only if it independently predicts held-out transitions.cases/projects/terraform-opentofu.events.yaml uses schema_version: 1, case_id: terraform-opentofu, coding status exploratory_seed, and reviewer_blinded_to_outcome: false. Its note says the inventory was assembled after inspecting pilot outcomes and may support exploration/report design but is not a blinded confirmatory event timeline.terraform-mitchell-board-exit-2021: institutional board_or_steering event for Terraform, titled Mitchell Hashimoto leaves HashiCorp executive team and board; affected scope is HashiCorp leadership/board and indirect Terraform control; power change is loss of executive-meeting access and formal board vote; confidence 1.0; primary HashiCorp source https://www.hashicorp.com/en/blog/mitchell-s-new-role-at-hashicorp. Announced and effective 2021-07-22. (meaning July 22, 2021)hashicorp-ipo-2021: institutional control_rights event for Terraform, titled HashiCorp becomes a publicly traded company; affected scope is ownership and capital-market accountability; public shareholders gain economic and formal voting rights, while detailed class rights require SEC coding; confidence 0.8; primary HashiCorp source https://www.hashicorp.com/en/blog/a-new-chapter-for-hashicorp. Announced and effective 2021-12-09. (meaning December 9, 2021)terraform-bsl-license-2023: institutional license event for Terraform, titled HashiCorp changes future Terraform releases from MPL 2.0 to BUSL 1.1; affected scope is source-code reuse and commercial competition; HashiCorp gains the right to exclude defined competitive commercial uses of future releases; confidence 1.0; primary HashiCorp source https://www.hashicorp.com/en/blog/hashicorp-adopts-business-source-license. Announced and effective 2023-08-10. (meaning August 10, 2023)opentofu-public-fork-2023: institutional fork event for OpenTofu, titled OpenTofu repository becomes publicly available; affected scope is Terraform-compatible open-source development; a separately governed MPL-licensed development path becomes publicly contestable; confidence 1.0; primary OpenTofu source https://opentofu.org/blog/the-opentofu-fork-is-now-available/. Announced and effective 2023-09-05. (meaning September 5, 2023)opentofu-linux-foundation-2023: institutional foundation_transfer event for OpenTofu, titled Linux Foundation launches OpenTofu under neutral governance; affected scope is project stewardship and cross-company governance; stewardship is assigned to a neutral foundation rather than one sponsoring vendor; confidence 1.0; primary Linux Foundation source https://www.linuxfoundation.org/press/announcing-opentofu. Announced and effective 2023-09-20. (meaning September 20, 2023)terraform-mitchell-departure-2023: institutional leadership event for Terraform, titled Mitchell Hashimoto leaves HashiCorp employment; affected scope is founder affiliation, with no remaining executive or board right reported; no formal right change was identified because those rights ended in 2021; confidence 0.9; primary HashiCorp source https://www.hashicorp.com/en/blog/mitchell-reflects-as-he-departs-hashicorp. Announced and effective 2023-12-14. (meaning December 14, 2023)opentofu-ga-2024: control-role product event for OpenTofu, titled OpenTofu 1.6 reaches general availability; affected scope is product maturity and adoption; no institutional right change, retained as a product-stage control; confidence 1.0; primary OpenTofu source https://opentofu.org/blog/opentofu-is-going-ga/. Announced and effective 2024-01-10. (meaning January 10, 2024)hashicorp-ibm-acquisition-2024-2025: institutional control_rights event for Terraform, titled IBM agrees to acquire HashiCorp and later completes the transaction; affected scope is ultimate ownership of HashiCorp and Terraform; IBM gains ultimate corporate appointment, budget, and strategy rights at closing; confidence 1.0; primary IBM announcement source https://newsroom.ibm.com/2024-04-24-IBM-to-Acquire-HashiCorp-Inc-Creating-a-Comprehensive-End-to-End-Hybrid-Cloud-Platform. Announced 2024-04-24. (meaning April 24, 2024)hashicorp-ibm-acquisition-2024-2025 became effective at transaction completion, with primary IBM completion source https://newsroom.ibm.com/2025-02-27-ibm-completes-acquisition-of-hashicorp,-creates-comprehensive,-end-to-end-hybrid-cloud-platform. (meaning February 27, 2025)opentofu-1-7-2024: control-role product event for OpenTofu, titled OpenTofu 1.7 ships state encryption and provider-defined functions; affected scope is product functionality and issue demand; no institutional right change, retained as a release-cycle control; confidence 1.0; primary OpenTofu source https://opentofu.org/blog/opentofu-1-7-0/. Announced and effective 2024-04-30. (meaning April 30, 2024)opentofu-1-8-2024: control-role product event for OpenTofu, titled OpenTofu 1.8 ships early evaluation and provider mocking; affected scope is product functionality and issue demand; no institutional right change, retained as a release-cycle control; confidence 1.0; primary OpenTofu source https://opentofu.org/blog/opentofu-1-8-0/. Announced and effective 2024-07-29. (meaning July 29, 2024)opentofu-cncf-sandbox-2025: institutional foundation_transfer event for OpenTofu, titled OpenTofu is accepted into the CNCF Sandbox; affected scope is foundation oversight and cloud-native project governance; the project enters CNCF’s staged project-governance and oversight framework; confidence 1.0; primary Cloud Native Computing Foundation source https://www.cncf.io/projects/opentofu/. Announced and effective 2025-04-23. (meaning April 23, 2025)opentofu-1-10-2025: control-role product event for OpenTofu, titled OpenTofu 1.10 ships as downloads approach ten million; affected scope is product functionality and adoption milestone; no institutional right change, retained as a product and adoption control; confidence 1.0; primary OpenTofu source https://opentofu.org/blog/opentofu-1-10-0/. Announced and effective 2025-06-23. (meaning June 23, 2025)research/session-handover-2026-08-27.md describes the repository as a reproducible lab testing whether institutional-power changes precede project-performance changes. Its authoritative boundary is exactly 3 raw statistical transition candidates, 0 robust transitions, and 0 canonical institutional-power-ledger records; the 40-record corpus and exploratory inventory still await source-level adjudication.main is authoritative, and the Gemini policy was frozen in commit 58a9663b8a44cf2d7bbb89102f5c70c00cdf0d91. The prior checkout was materialized from remote files over a synthetic local baseline, so many modified/untracked files may be legitimate upstream content; the handover warns against resetting or bulk-publishing that status and requires exact-file comparison with remote main.60.0% of 10, 36.7% of 30, and 46.7% of 30 eligible specifications, all during OpenTofu’s first year; therefore 3 raw candidates yielded 0 robust transitions, with startup/backlog normalization preferred over narrative fitting.79bde4d6c5cc187231c6108b8439c58cdb26b220 (79bde4d) — cursor pagination and regression test; 2. 1d3af619e018f3e5050152c847cd5162219375a9 (1d3af61) — robustness analysis, event timeline, and report; 3. 2741626670eeee678948916c7a237b1c483f88a6 (2741626) — bounded retries for dropped HTTP responses.github.token with copilot-requests: write and requires the exact requested model, with no fallback. Model sessions disable shell, MCP, custom tools, repository writes, skills, config discovery, memory, and session storage; each document is isolated and only allowlisted fields enter prompts.100%, institutional recall 90%, event-kind accuracy 96%, evidence grounding 100%, and variable strict power-edge recall from 75% to 91.7%.75% institutional recall and all missed Valkey; higher reasoning increased latency without improving discovery. Luna/medium had the best mean strict-edge recall among the four, but none justified automatic graph serialization.cases/model-selection/event-triage-v1.json: research role Event candidate discovery and triage only; model gemini-3.7-flash; provider-default reasoning with SDK omission of reasoningEffort; selection run 33086454444; artifact 9652609349; digest sha256:44ea0132eb7416e2c4ad530e66b4e5c672700fd0eb2d9db6be5e5cb196bcd2d0.gemini-3.7-flash strict selection results were 95.8% decision accuracy; 100% institutional precision, recall, kind, date, and minimum evidence grounding; 3/3 Valkey recovery; 75% strict-edge recall; and mean latency 8,068 ms/document on the repeated hard slice.gemini-3.7-flash was chosen under the cheapest-passing stop rule; mai-code-1.1-flash failed with 75% decision accuracy, 75% institutional recall, and 0/3 Valkey recovery; Sonnet 5 and Kimi K3 were not called because Gemini triggered the stop rule; Sol was excluded and must not be described as an empirical failure; Raptor mini was unavailable in the Copilot runtime.33063470968; artifact 9642983578; 40-record document bundle SHA-256 d70f7f795847968339a645b6dbd6d4428af6bd6044de44fa2602d4c4343b0ee9; 5/40 provisional Luna core agreements; 35/40 routed to adjudication; 0 canonical-ledger records.33086454759, artifact 9653837747, digest sha256:65dba871145d26feab8eb5938a0e9ef44dc36c95215b207549d95cfe9691a6b9; it produced 8/40 provisional core agreements and 32/40 routed records but worse strict-edge agreement. It is stability evidence only and does not replace frozen V1.2.33108057720, successful; 2. explicit Luna compatibility run 33108057688, successful, artifact 9661404982, digest sha256:7a6a80fcd5ff5eb0db24a955827071f827fc2fea89723fecd3cd76077eb8afc4; 3. full retrieval/coding run 33108057787, successful, artifact 9662866867, digest sha256:aa598eac82a2c4d6de072ca19e0c59291730fded4203a98d7b6a91f4ff9f59c8.gemini-3.7-flash, provider-default reasoning, policy ID event-triage-cost-frontier-v1, and only 4 allowlisted input fields. Results were 224 schema-valid responses (99.56%), 4 institutional candidates, and a 4-record review queue; 1 response altered an evidence quotation and was quarantined, while accepted evidence remained 100% grounded and integrity passed.d70f7f795847968339a645b6dbd6d4428af6bd6044de44fa2602d4c4343b0ee9; two explicit Luna coders, Terra advisory routing, 35 adjudication records, 5 provisional core agreements, and 0 canonical-ledger records verified that the triage-default change did not alter coding roles or bypass adjudication..github/workflows/retrieve-oss-governance.yml performs retrieval, selected-model triage, fail-safe enrichment, frozen blinded coding, Terra routing, and artifact upload; changes to shared extraction code may intentionally trigger this expensive end-to-end workflow..github/workflows/compare-sol-reasoning.yml retains a legacy filename but implements the cost-frontier protocol and is manual-only after the completed experiment. .github/workflows/probe-luna-events.yml remains the historical 25-document Luna benchmark and explicitly requests Luna rather than inheriting the Gemini default.33060493337 on main at commit 9ba7d77baec0b001649a4d455330a9fe298a05e9 produced unexpired artifact 9642534015, named oss-governance-retrieval-v2-9ba7d77baec0b001649a4d455330a9fe298a05e9, size 1,388,304 bytes, digest sha256:d3a84341050ac9f6de3f55ac7d22fd83c283fa7910eecd6fa7172ed323b4d80f. Created 2026-08-27 and expiring 2026-09-26. (meaning September 26, 2026)33053908119 on main at commit 3526b683c16fb130545034c3dbb795d0b4fca6c8 produced unexpired artifact 9639245355, named oss-governance-retrieval-v2-3526b683c16fb130545034c3dbb795d0b4fca6c8, size 1,114,479 bytes, digest sha256:0579bcb69136e0f8ed42d9b002aaab21ccb77046010d6263f5e30c45114c812d. Created 2026-08-27 and expiring 2026-09-26. (meaning September 26, 2026)