Repository navigation
Conversation
…SIM-2/3, Refs #5511) Agent: aprender-w5404 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…erge queue, ledger inputs, sim block (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…w/repo-flow/course-flow (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…-input overrides (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…easured inputs and sets drift_pct_24h (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…tle's law, same bytes); pin pr-flow in the surface gate (Refs #5511) Mutant r_mq 4->1 on design D reads 7.36 vs 20.08 and the test goes red; reverted. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…tual, one_change, optimum), one adapter, test first (Refs #5511) Params shape is provisional until infra pins arbiter-sim-params-v1; only parse_params knows it. p_dirty and t_rebase are accepted and reported as unmodelled until SIM-6. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ithin 0.5% (RED; flow model B6, Refs #5545) Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…o utilisation thresholds (flow model eq 2,3,5,6) (Refs #5511) Mutation: min_by -> max_by turns 3 binding tests RED; restored GREEN. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…d reports its mean (flow model B6, Refs #5545) batch_in_wip (opt-in; reference bands unchanged), mean_batch_wait_h in the result and sim block. Mutation: dropping the batch-time slot release turns littles_law_holds_with_batch_wait_inside_wip RED. Also pins the pr/repo/course-flow subcommands in the clap reachability table (12). Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…n batch gate inside WIP (flow model eq 1, Refs #5545) Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… pr-flow to subcommand table (flow model eq 7) (Refs #5511) Mutation: dropping t_rev from the arm delay -> dL 0, test RED; restored GREEN. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…e 0.95 assumption (flow model eq 2) (Refs #5511) Mutation: zeroing the blocked-time term makes the capped case read 1.0, test RED; restored GREEN. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…checks, reviewer lanes, per-tier review minutes (flow model eq 7) (Refs #5511) Falsifier: concurrent review with t_rev <= check wall saves the serial t_rev within 5%. Mutations: concurrent arming disabled -> RED; lane release dropped -> RED; restored GREEN. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ue wait is the pass in flight, not the batch timer (Refs #5511) Root cause of sim W below law W on design D: the law took round inputs (w=5, c_wall=30, pass=30) and used the 5-minute batch timer as the merge queue wait. With one build at a time the build slot is busy, so an armed PR waits out the pass in flight plus any overflow past the batch size (17.2 min, not 5). The checks term ran 9.7 min high and the queue term 7.8 min low; they partly cancelled into a plausible gap. This line has no rebase path, so p_dirty = 0 and rebase cannot explain it. The sim now reports stage times (run wall, open->green, green->armed, armed->merged, wait for first pass, pass wall) and the --json block gains a "law" object: eq 7 evaluated on the design's own config with the queue waits read from the run, term by term, beside predicted.W_min_mean. Design D: sim 82.0 vs law 82.0 min (0.05%); G 0.12%, E 0.10%. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…i-commands-v1; regen contracts.nt and shapes.ttl with the repo-built pv (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…CHEMA-026); regen ONT witness and README contract count (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…003 (Refs #5511) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…q 6 law, build at eq 14 law (RED; flow model B5, Refs #5557) Fan-out to R = 3 renditions with no split (eq 3), the eq 6 stage law 5 * 360 / (1440 * 2) = 62.5% within 2 points, alpha rework at 0.16 keeping build busy at 79.5% / (1 - alpha) <= 95% (eq 14), GPU load B*t_g + (B*k + 4)*t_v (eq 5), trunk-first GPU priority while W > 4.8 d, and t_b work (eq 2) split from t_b elapsed (eq 1). RED: the v2 config and result fields do not exist yet. Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> (cherry picked from commit 2985833)
…priority, build work vs elapsed (flow model B5, Refs #5557) Opt-in config fields; the defaults reproduce the original model, so every existing band test is unchanged (same RNG draws, same event stream). - fan_out: every master fans out to R = 3 renditions with no split (eq 3); launches_per_day = 3 * courses_per_day. - rework_alpha: the distinctness gate sends a share alpha back to build (eq 14). - trunk_first_w_days: the GPU serves the trunk first while the recent course W (last 20 completions) exceeds it, else Shorts. - linkedin_renders_per_day, vertical_render_min: feed renders share the GPU (eq 5). - build_elapsed_min, and result t_b_work_h / t_b_elapsed_h: build work (eq 2, sizing) split from build elapsed (eq 1, lead time). Ledger inputs t_b_work_h and t_b_elapsed_h; both, and stage_util_pct / launches_day, appear in the sim block. - Preset factory-v2 keeps the 07:00 UTC batch, counted inside WIP. Measured (seeds 1..5, 300 d, 20 d warm-up): stage 62.47% vs eq 6 law 62.5%; legacy split 24.97% (= 0.4 of the law); build at alpha 0.16 94.27% vs eq 14 law 94.64%; GPU 37.6% vs eq 5 law 37.8%. Finding: the GPU priority rule reorders work and cannot add GPU minutes. Shorts are never dropped, so under GPU overload courses/day is the same either way and Shorts carry the wait. The RED test that asserted +0.08 courses/day was replaced by a mechanism test of the pick and a conservation test. Mutation: rework disabled -> rework_at_alpha_016 RED (0.793 vs 0.947); priority disabled -> gpu_serves_trunk_first... and gpu_priority_conserves... RED; fan-out reverted -> coursera_stage_matches_the_law RED (0.252 vs 0.625). Restored: green. Contract: contracts/course-flow-content-factory-v2-v1.yaml (pv validate: valid). Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> (cherry picked from commit 9469789)
…-content-factory-v2 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
§13.11 rung 1 — quorum shadow verdict Shadow mode: this records a verdict and merges nothing. A refusal |
noahgift
marked this pull request as draft
October 18, 2026 18:19
…nt-factory-v2 census The B5 contracts changed the census but the witness still recorded the old one, so ont5_consistency_gate the_spec_probe_holds_on_the_repo_corpus failed with WitnessStale. Output of scripts/contracts_gate.sh (pv-sat). Agent: aprender-w5032 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… squashed; keep B5 course_flow, regen README/nt/witness) (Refs #5557) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Contributor
Author
|
Closing under the P7 WIP freeze (open ≤10): stale. Its base #5559 merged at 02:23Z, so this stacked draft is CONFLICTING and has had no seats. Branch batch/b5-content-factory-on-5559 is kept; reopen as a fresh PR rebased on main once the freeze lifts. |
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.
STACKED on #5559 (head 0ecee41): this branch is #5559's head plus two cherry-picked commits (2985833, 9469789: flow model B5, Content Factory v2) and a derived-file regen. Do not arm until #5559 has merged; after that the diff against main is only the B5 commits.
Refs #5557 #5545
🤖 Generated with Claude Code