Skip to content

Latest commit

 

History

History
788 lines (769 loc) · 133 KB

defects.md

File metadata and controls

788 lines (769 loc) · 133 KB

Defect log

523 defects recorded across the packing toolchain. One line each here; the narrative lives in the artifact named by every row. Source of truth is defects.yaml.

The short version

  • 109 soundness defects — the system asserting something false about the mathematics. 85 of them pointed in the flattering direction, which is the dangerous one: the error looks like success.
  • 121 fixes left no regression check behind. That list is the best predictor of what comes back; recorded recurrences are D-017 repeats D-010, D-029 repeats D-023, D-062 repeats D-042, D-065 repeats D-028, D-066 repeats D-042, D-072 repeats D-035, D-075 repeats D-059, D-076 repeats D-034, D-077 repeats D-028, D-078 repeats D-041, D-079 repeats D-063, D-082 repeats D-057, D-085 repeats D-058, D-094 repeats D-084, D-098 repeats D-083, D-104 repeats D-052, D-113 repeats D-100, D-115 repeats D-097, D-117 repeats D-104, D-138 repeats D-006, D-140 repeats D-093, D-148 repeats D-091, D-150 repeats D-086, D-155 repeats D-059, D-160 repeats D-145, D-162 repeats D-030, D-163 repeats D-004, D-164 repeats D-014, D-165 repeats D-132, D-166 repeats D-044, D-168 repeats D-132, D-169 repeats D-014, D-170 repeats D-039, D-171 repeats D-164, D-172 repeats D-029, D-180 repeats D-086, D-181 repeats D-034, D-187 repeats D-185, D-188 repeats D-018, D-189 repeats D-181, D-196 repeats D-160, D-198 repeats D-187, D-201 repeats D-198, D-204 repeats D-201, D-217 repeats D-202, D-229 repeats D-028, D-242 repeats D-232, D-247 repeats D-242, D-255 repeats D-198, D-259 repeats D-027, D-263 repeats D-258, D-267 repeats D-255, D-274 repeats D-268, D-279 repeats D-271, D-281 repeats D-267, D-282 repeats D-264, D-312 repeats D-309, D-313 repeats D-259, D-315 repeats D-295, D-318 repeats D-308, D-321 repeats D-317, D-323 repeats D-022, D-324 repeats D-320, D-325 repeats D-319, D-326 repeats D-305, D-327 repeats D-301, D-334 repeats D-028, D-337 repeats D-107, D-339 repeats D-155, D-340 repeats D-163, D-386 repeats D-358, D-395 repeats D-358, D-397 repeats D-358, D-400 repeats D-398, D-422 repeats D-371, D-424 repeats D-144, D-427 repeats D-413, D-451 repeats D-442, D-452 repeats D-442, D-454 repeats D-442, D-458 repeats D-443, D-470 repeats D-458, D-471 repeats D-459, D-488 repeats D-432, D-502 repeats D-488, D-509 repeats D-490, D-517 repeats D-453.
  • 73 are still open (outstanding or contained), every one carrying a bead.

What caught them

Detector Count What it is
pre_registered_rule 3 a rule written down before the measurement, e.g. “beating the record means you have a bug”
control_cell 29 a cell of the sweep whose answer is known in advance
review 306 a human or agent reading the work against a checklist
anomaly 16 a result that made no sense, chased down
inspection 68 reading the code or the design with intent
drift_check 17 a generated view disagreeing with its source
design 1 caught while designing, before it reached data
gate 83 the automated test suite

The line worth reading twice: the automated gate caught 83 of 523, and none of the 109 soundness defects. Gates confirm what you already thought to check. The rest were found by a device built to be surprised — a control cell, a pre-registered rule, a generated view contradicting itself — or by someone reading carefully.

Where they arise

Layer Count
engine 11
quench 23
verifier 15
record 162
tooling 196
docs 116

By kind

Class Count
soundness 109
validity 128
bookkeeping 196
robustness 71
performance 19

Fixed, but nothing stops it coming back

Each of these had its cause removed and no check added. This is the actionable list.

id title layer
D-002 A restart cap made the declared move budget inert engine
D-007 The bead tree never reached the shared store record
D-010 A three-cell sweep recorded as a one-cell round record
D-015 The angle objective was not a function of the angles quench
D-016 One variable served as both the finite-difference probe and the line-search step quench
D-017 A three-cell sweep recorded as a one-cell round, again record
D-026 The plan spec's checklist contradicted the state the spec was cited as authority for record
D-032 The in-progress stub was not schema-valid, so the recovery path corrupted the record record
D-033 The ledger crashed on any in-progress round, so the gate could not run during a session record
D-038 Closed-form recognition was documented as a local-optimum oracle tooling
D-055 An algebraic trigonometric coordinate was used to call the angle algebraic docs
D-056 Grid optimality for m squared minus 3 was generalized beyond the proved cases docs
D-057 One n=17 regime was generalized to oblique-search blindness at every n docs
D-063 A non-converse was called the contrapositive of the rigidity premise docs
D-066 The active baseline script still classified open n=12 as a negative control docs
D-069 H-002 still said its measured quench instrument was not built record
D-075 The PR 16 response claimed a portable oracle from an aggregate byte mismatch docs
D-076 Six distinct n=5 endpoint rows were treated as proof that identity was too fine docs
D-077 The PR 16 handoff presented stale state and priorities as a current dependency queue record
D-078 The n=5 rank correction still named equality independence as the sufficient condition docs
D-079 The handoff retained an invalid implication from record rigidity to non-record behavior docs
D-080 The original neighbor-transfer criterion was satisfied by the cold-start grid docs
D-082 The synopsis again generalized one n=17 regime to oblique-search blindness docs
D-083 A finite-quench residual was called evidence of attraction through epsilon 0.1 docs
D-084 The n=11 frontier artifact asserted rigidity and the wrong gap rank docs
D-085 Living run commands could rewrite the locked dependency state tooling
D-086 The overnight agenda and current handoff retained obsolete work and effort state record
D-087 Angle-class evidence and an untested kink sweep combined distinct claims docs
D-088 The finite-n frontier missed newer certified n=68 and n=69 constructions docs
D-089 The n=17 record was given symmetric plus-or-minus 40-degree orientations docs
D-090 Two low-angle examples were generalized into a universal tiny-angle prior docs
D-091 H-010 did not reproduce Stromquist's conditional proof mechanism docs
D-092 Structured asymptotic data repeated the phantom Roth-Vaughan constant docs
D-094 The idea board repeated the wrong n=11 gap rank docs
D-095 The idea board called a refuted hypothesis the best runnable experiment record
D-096 Algebraic degree was promoted into a proof-method ceiling and obstruction docs
D-097 Strategy catalogues were described as exhaustive histories of the field docs
D-098 Source code still called a finite stochastic return threshold a basin radius engine
D-099 H-023 placed six non-optimal endpoints in one optimum-side family question docs
D-100 Delta continuation had no preserved parameter and reversed its progress metric docs
D-102 H-006 treated dual pose support as candidate piercing-point loci docs
D-103 H-011 omitted its declared component-identifiability prerequisites record
D-104 Living artifacts still promoted local refinement failure into a wrong-basin claim docs
D-105 Superdisk continuation lacked a shape normalization and branch rules docs
D-108 A direct piercing-number application was missing from the method history docs
D-109 Boundary corner coordinates were double-counted as independent wall equations docs
D-110 A fixed-cell LP dual was mislabeled a free-framework self-stress docs
D-111 A fixed-budget extreme-value fit was generalized to unlimited search support docs
D-112 One fixed-cell angle sheet was presented as the global two-class landscape docs
D-113 Numerical branch coalescence was called a feasible-component merge and barrier docs
D-114 Fractional piercing was given an integral and symmetric conclusion docs
D-115 Negative novelty claims exceeded the searched literature surface docs
D-117 The idea board again said a basin-width question had been answered docs
D-119 A continuous angle sheet was required to have only one nearby grid point record
D-127 The engineering review called its branch complete before its delta blockers were resolved docs
D-146 The Stromquist Figure 13 transcription changed its defining point set docs
D-147 H-010 substituted unit squares for Stromquist's strict open boxes docs
D-148 The D-091 correction omitted the published Figure 14 unavoidability claim docs
D-149 Search saturation was still presented as a Stromquist known-answer test docs
D-160 A broad H-010 integration patch updated D-002 instead of D-151 record
D-172 The orientation tutorial said a configuration-space cell fixes the angles docs
D-173 The orientation docs ruled out equality certification by all interval methods docs
D-174 The recovery narrative silently eliminated centres and universalized rational LPs docs
D-176 The orientation tutorial promoted special-case angle evidence into general structure docs
D-177 The orientation docs asserted unmeasured continuity and language-independent cost docs
D-179 The orientation tutorial treated undefined hardness and s(11) tractability as settled docs
D-181 The basin definition called a deterministic quench endpoint a non-point docs
D-182 The strategy tutorial excluded safe loss shaping and all repeated endpoint hits docs
D-189 The newcomer glossary omitted the point-basin and component boundary docs
D-191 Orientation prose overstated both control automation and conventions ownership docs
D-193 The draft PR 21 review misclassified a valid but incomplete definition as soundness docs
D-207 The workflow map handed repairs to an undefined implementation workflow docs
D-208 The new campaign definition excluded work already stored in the campaign docs
D-209 The legacy-series summary stopped at 35 of 36 experiments docs
D-211 The workflow cadence mandated review phases after routine guarded rounds docs
D-212 The workflow prose turned a primary focus into an exclusive focus docs
D-216 Full phase histories made the generated orientation table unboundedly wide tooling
D-218 Session-009's output inventory stopped before exp-035 and exp-036 record
D-219 The BC-010 agenda retained a superseded two-size solver diagnosis record
D-220 The durable resume order was not encoded in the ready-work graph record
D-221 The campaign bead retained the superseded eight-hour horizon record
D-222 A delegated read-only audit launched an explicitly excluded deep gate record
D-223 The synopsis retained superseded normal-gate timing and control counts docs
D-224 A focused negative-control replay omitted the gate-managed Python environment record
D-225 Checkpoint merge readiness was conflated with unattended launch readiness record
D-230 Top-level synopsis claims exceeded the scope of the detailed lane evidence record
D-232 An unfrozen lint command rewrote the packing dependency lockfile tooling
D-234 The implemented engineering plan carried stale inventory and defect references docs
D-236 A completed timeout fix remained outstanding in the defect and launch records record
D-238 The synopsis called quench endpoints discrete basins with a well-defined census docs
D-243 The autonomous-loop instruction contradicted its upstream launch blocker docs
D-247 A delegated follow-up again rewrote the dependency lock outside its scope tooling
D-280 The phase-count cap ended a fast campaign before its wall-clock goal tooling
D-284 A delegated float scout ran before the Trump support criterion was frozen record
D-287 A shell-active bead description executed a backtick-delimited word record
D-289 The H-043 pilot normalized row classes for counting but not for its cone oracle tooling
D-291 The branch-0 H-043 golden would classify a valid future refutation as test failure tooling
D-292 H-043 branch indices are regenerated but not bound to the retained exp-013 universe record
D-297 The first Windows timeout path treated taskkill as a proved tree guarantee tooling
D-298 Overlapping inspection ranges fabricated a duplicate session key tooling
D-299 A later successful check masked an earlier gate failure in the shell status tooling
D-301 The first receipt rule split a compound adjective after its hyphen docs
D-302 Bui transcription changed the square-reduction divisor into a product docs
D-303 Bui transcription omitted the quantified width hypothesis docs
D-304 McClenagan's Section 3 proof uses a contradictory inequality chain docs
D-310 McClenagan's equation 3.2 calls a periodic tangent solution unique docs
D-322 The synopsis overstated which defect classes the gate had caught docs
D-327 The timeout documentation split another compound adjective after its hyphen docs
D-474 Feasible covering masses were subtracted to infer a lower slope and a stopping side record
D-475 The README presented a fixed-net numerical ceiling without its net restriction docs
D-478 The BC-206 cutting floor has no retained family for independent replay record
D-479 Case summaries replaced cited geometric helper proofs with generic pure dots counting docs
D-481 The n26–27 source fields omitted Green's stronger reported lower bound record
D-495 Unfinished covering and interval checks were treated as refutations record
D-503 Artifact timeout test assumed a Python child could print within 200 milliseconds tooling
D-504 Allowlist existence test generated a full declared-bound reference report tooling
D-505 Bentz 2016 transcription moved the finishing line of Theorems 9 and 11 docs
D-506 Bentz 2016 transcription inverted Lemma 7's bound and rewrote its lemma reference docs
D-508 Rung-0 records counted the reader's branch nodes as leaf certificates record
D-514 The Erdős–Graham transcription stated Theorem (1) as a two-sided bound the paper does not prove docs
D-515 The McClenagan transcription swapped Montgomery's exponent, and the record then faulted McClenagan for it docs

Still open

id status severity title bead
D-021 contained medium Floating LP refinement has a noise floor, and eight rounds sit on it think-hg3u
D-034 outstanding high Endpoint identity assumes isolated terminals and splits a connected optimum family think-1s0h
D-039 outstanding high A scalar side-error floor was treated as a basin-resolution theorem think-3szr
D-040 outstanding high Basin rarity was reported without a proposer-conditioned estimand or durable sample think-apwt
D-041 outstanding high Raw contact counts and a one-dimensional kink were used as rigidity and dimension proofs think-1s0h
D-044 contained critical Result validity and self-test status are assertions from the proposer itself think-ldq2
D-045 outstanding high The generic criterion evaluator silently reduces unlike hypotheses to side gap think-rrht
D-046 contained critical The unattended runner is not a closed, checked state machine think-ldq2
D-048 outstanding high Tolerance grouping and exact hash pairs do not form a stable basin equivalence relation think-siui
D-049 outstanding high Contact-graph canonicalization is factorial on sparse symmetric endpoints think-siui
D-050 outstanding critical The atlas promotes non-converged stopping points and cannot reconstruct discovery order think-31k1
D-051 outstanding high Atlas frequencies merge without regime, shard, schema, or identity provenance think-31k1
D-052 outstanding high Coordinatewise quench stopping was reported as a certified local optimum think-zcx4
D-054 outstanding high Move budgets overshoot and final-best records cannot support trajectory claims think-rrht
D-059 outstanding critical The golden map mixed unhermetic discovery snapshots with mathematical pass-fail assertions think-zt29
D-061 outstanding high Unrecognised endpoint rows discard the evidence needed to classify them think-aans
D-071 outstanding medium Generated session reports overwrite history and are not durable think-y37w
D-081 outstanding high Runner preflight treats any nonempty queue as enough unattended work think-kmn2
D-101 outstanding high Historical quench round wall times disagree with retained per-call durations think-xzew
D-126 outstanding high Quench convergence work was budgeted by machine time rather than reproducible work think-u97a
D-139 contained high H-032 omitted directly relevant hard-square configuration-space literature think-izep
D-162 outstanding high The small-n golden convergence labels hid unsettled fixed-cell evaluations think-wbra
D-202 contained medium A delegated long-running validation lost its final receipt think-b3bm
D-217 contained medium A parallel validation wrapper discarded a live command receipt think-b3bm
D-222 contained medium A delegated read-only audit launched an explicitly excluded deep gate think-ysz2
D-239 outstanding medium Validation steps have no outer deadline or process-group cleanup think-tx0b
D-280 contained high The phase-count cap ended a fast campaign before its wall-clock goal think-bfwm
D-283 outstanding high A zero-step sqsearch regime can spin forever without spending its move budget think-dprg
D-289 contained high The H-043 pilot normalized row classes for counting but not for its cone oracle think-jyit
D-290 outstanding medium H-043 pilot certificates can be recomputed but not replayed from a retained record think-oa96
D-291 contained high The branch-0 H-043 golden would classify a valid future refutation as test failure think-j92q
D-292 contained high H-043 branch indices are regenerated but not bound to the retained exp-013 universe think-7jyh
D-293 outstanding medium Two H-043 provenance selftests claim completeness without checking exact mappings think-8wgw
D-300 outstanding medium The receipt rehearsal requested an unsupported gdate precision think-jygr
D-304 contained high McClenagan's Section 3 proof uses a contradictory inequality chain think-486e
D-343 contained medium Bui's replacement instruction omitted its finite upper index bound think-ykt7
D-344 contained high El Moumni's n=7 proof prints a negative segment length think-trkj
D-345 contained high El Moumni's n=7 theorem drops Proposition 2's minimum branch think-trkj
D-346 contained high El Moumni's Figure 4 chord line passes through the wrong square center think-trkj
D-347 contained high El Moumni's Figure 4 diameter argument uses an undefined point i think-trkj
D-349 contained low An aborted free sweep drops its LP work from the run counter think-7dwo
D-356 outstanding low The negative-control snapshot prunes the archive, so the n=29 chain cannot be controlled there think-lo3p
D-357 outstanding low One synopsis negative control failed reproducibly, then passed, with no identified trigger think-lo3p
D-362 outstanding low The translation-only trajectory check compares independently rounded projections for exact equality think-mt4h
D-365 outstanding medium The deep golden-basin oracle fails at n = 10, and only the strict tier runs it think-c7oo
D-377 contained low A read-only delegation cannot be recorded while it is running think-s424
D-378 contained medium No retained control can separate the surviving identity relation from merging everything think-6zaz
D-391 outstanding high A tangent cone that is a union of half-spaces was intersected think-s424
D-393 contained medium The pre-push floor does not run tests, and CI stayed red for seventy-five minutes think-s424
D-403 contained medium The negative controls do not run on a pull request, so stale ones accumulate unseen think-xdly
D-405 contained medium A sequencing plan was built on an unmeasured claim that the verified gap ranks nothing think-bxqv
D-406 outstanding medium BC-092 was stopped on an enumeration price that traces to no artifact think-bxqv
D-418 outstanding high The BC-140 bound checker passed eight bounds with no named exceeding mutation think-ifgr
D-419 outstanding high Exp-057 applied six-decimal coordinate semantics to a fourteen-digit side token think-lvqx
D-420 outstanding high The BC-142 reachable-tests control did not prove exact selection equivalence think-mo7r
D-421 outstanding medium The wave-efficiency renderer refused the mixed Claude-to-Codex agenda lanes think-mlwo
D-423 outstanding medium The cover certificate's Lemma 4 boundary is non-strict in two sources and strict in a third, unrecorded think-iye2
D-426 outstanding high A control-cell breach does not stop the unattended session think-ldq2
D-427 outstanding medium Controls are scored on the mutated run alone, so one can pass over an already-red checker think-g4qi
D-428 outstanding medium The n = 17 successor validator never ties the rebuilt chain spine to the carried boundary think-g4qi
D-429 outstanding low The generated-view comparison folds an ellipsis but not the space the formatter puts before it think-ahyr
D-431 outstanding medium T-009's significance rationale compares its interval certificate against a different packing think-stb5
D-460 contained low A research cell with no hypothesis cannot produce a round, and BC-206 was written that way think-c0qm
D-467 outstanding medium Two branches allocated the same three defect ids to different defects, and nothing local can see it think-8bcl
D-470 contained medium A test pinned to the moving certificate pointer went stale, and no pull request could have caught it think-doar
D-471 contained medium The conflicted-branch CI blackout fired three more times in one day, and nothing but a person detects it think-706h
D-472 outstanding medium A single CI timing was treated as a measurement, and a twenty per cent difference was attributed inside thirty-four per cent of noise think-be1s
D-478 contained medium The BC-206 cutting floor has no retained family for independent replay think-aenh
D-489 outstanding high The corner-dual salvage screen can only accept a family that already fails the mass threshold think-rm5c
D-490 contained medium Two renders of the explainer disagreed by two bytes and the check could not say where think-ptit
D-502 outstanding high Costly partial pre-push selections kept pytest serial until the command timed out think-1i1x
D-509 contained medium One math glyph's baseline is decided per page load, so drawing each load twice does not contain it think-6dle
D-521 outstanding medium Numerical witness receipts record tolerance-blurred values as the minimum clearance and pair gap think-jtot

Every defect

id date layer class dir caught by severity status title
D-001 2026-08-22 engine validity conservative control_cell high fixed Fixed-side search never left the grid basin
D-002 2026-08-22 engine validity neutral anomaly high fixed A restart cap made the declared move budget inert
D-003 2026-08-22 engine performance inspection medium fixed Squared overlap penalty has a vanishing gradient
D-004 2026-08-22 tooling validity flattering inspection high fixed Half the test suite was silently unreachable
D-005 2026-08-23 record bookkeeping review high fixed Campaign schemas declared enforced, and nothing enforced them
D-006 2026-08-23 record bookkeeping review high fixed The archived run could not regenerate its own configurations
D-007 2026-08-23 record bookkeeping review medium fixed The bead tree never reached the shared store
D-008 2026-08-23 record bookkeeping review medium fixed Renumbering left dead links and a script writing to a deleted path
D-009 2026-08-23 engine soundness neutral review high fixed The overlap guard was asserted against a drifting accumulator
D-010 2026-08-23 record bookkeeping review medium fixed A three-cell sweep recorded as a one-cell round
D-011 2026-08-23 engine soundness conservative control_cell high fixed The annealer never scored its own starting configuration
D-012 2026-08-23 engine robustness control_cell medium fixed A zero perturbation made the cooling schedule degenerate
D-013 2026-08-23 engine soundness flattering design high avoided An infeasible chain would have reported a perfect return
D-014 2026-08-23 quench soundness flattering pre_registered_rule critical fixed The LP solver returned a packing that violated its own constraints
D-015 2026-08-23 quench validity neutral anomaly high fixed The angle objective was not a function of the angles
D-016 2026-08-23 quench validity conservative inspection medium fixed One variable served as both the finite-difference probe and the line-search step
D-017 2026-08-23 record bookkeeping review medium fixed A three-cell sweep recorded as a one-cell round, again
D-018 2026-08-23 record bookkeeping drift_check high fixed One round, two hypotheses, one verdict
D-019 2026-08-23 quench robustness conservative anomaly medium fixed The bracketing quench does not always terminate
D-020 2026-08-23 quench soundness flattering inspection high fixed The quench's answer depends on a tuning parameter, and basin identity would inherit it
D-021 2026-08-23 quench soundness flattering inspection medium contained Floating LP refinement has a noise floor, and eight rounds sit on it
D-022 2026-08-23 record bookkeeping control_cell low fixed A numeric field and its display duplicate could disagree in silence
D-023 2026-08-23 tooling bookkeeping review medium fixed Verification was done in throwaway snippets and left nothing behind
D-024 2026-08-23 record bookkeeping gate low fixed A strategy's enum and its prose said opposite things
D-025 2026-08-23 record bookkeeping review medium fixed Two bead trees claimed the same spec phase after the plan was rebuilt
D-026 2026-08-23 record bookkeeping review medium fixed The plan spec's checklist contradicted the state the spec was cited as authority for
D-027 2026-08-23 tooling bookkeeping review medium fixed A generated view was left reflowable by the Markdown auto-formatter
D-028 2026-08-23 record bookkeeping drift_check low fixed A defect-log headline stayed asserted after it stopped being true
D-029 2026-08-23 quench validity conservative review medium fixed A single fixed-angle LP solve was called "the quench", and retracted a correct finding
D-030 2026-08-23 quench validity flattering control_cell high fixed The angle window narrowed on a schedule, so a cold quench could never arrive
D-031 2026-08-23 tooling soundness flattering control_cell high fixed Basin identity split an angle at the pi/2 seam, inflating counts near axis-aligned optima
D-032 2026-08-23 record bookkeeping review high fixed The in-progress stub was not schema-valid, so the recovery path corrupted the record
D-033 2026-08-23 record robustness review high fixed The ledger crashed on any in-progress round, so the gate could not run during a session
D-034 2026-08-23 tooling soundness flattering control_cell high outstanding Endpoint identity assumes isolated terminals and splits a connected optimum family
D-035 2026-08-23 tooling robustness anomaly high fixed An interrupted negative-control run leaves its deliberate sabotage in the working tree
D-036 2026-08-23 quench validity flattering review high fixed A timed-out free sweep was reported as a convergence certificate
D-037 2026-08-23 tooling bookkeeping review low fixed The atlas checker reported its own synthetic re-offers as census proposals
D-038 2026-08-23 tooling soundness flattering review high fixed Closed-form recognition was documented as a local-optimum oracle
D-039 2026-08-23 tooling soundness flattering review high outstanding A scalar side-error floor was treated as a basin-resolution theorem
D-040 2026-08-23 record validity flattering review high outstanding Basin rarity was reported without a proposer-conditioned estimand or durable sample
D-041 2026-08-23 docs soundness flattering review high outstanding Raw contact counts and a one-dimensional kink were used as rigidity and dimension proofs
D-042 2026-08-23 record validity flattering review high fixed An open n = 12 case was used as a known-answer negative control
D-043 2026-08-23 record soundness flattering review high fixed Runner validation happened after archival and could be bypassed during record
D-044 2026-08-23 record soundness flattering review critical contained Result validity and self-test status are assertions from the proposer itself
D-045 2026-08-23 record validity flattering review high outstanding The generic criterion evaluator silently reduces unlike hypotheses to side gap
D-046 2026-08-23 record robustness review critical contained The unattended runner is not a closed, checked state machine
D-047 2026-08-23 tooling soundness flattering review high fixed Contact certificates changed under container reflection while the D4 test checked only geometry
D-048 2026-08-23 tooling soundness flattering review high outstanding Tolerance grouping and exact hash pairs do not form a stable basin equivalence relation
D-049 2026-08-23 tooling performance review high outstanding Contact-graph canonicalization is factorial on sparse symmetric endpoints
D-050 2026-08-23 tooling soundness flattering review critical outstanding The atlas promotes non-converged stopping points and cannot reconstruct discovery order
D-051 2026-08-23 record validity flattering review high outstanding Atlas frequencies merge without regime, shard, schema, or identity provenance
D-052 2026-08-23 quench soundness flattering review high outstanding Coordinatewise quench stopping was reported as a certified local optimum
D-053 2026-08-23 verifier soundness neutral review critical fixed The generic number-field API accepts reducible or multi-root certificates
D-054 2026-08-23 engine validity flattering review high outstanding Move budgets overshoot and final-best records cannot support trajectory claims
D-055 2026-08-23 docs soundness flattering review high fixed An algebraic trigonometric coordinate was used to call the angle algebraic
D-056 2026-08-23 docs soundness flattering review high fixed Grid optimality for m squared minus 3 was generalized beyond the proved cases
D-057 2026-08-23 docs validity conservative review high fixed One n=17 regime was generalized to oblique-search blindness at every n
D-058 2026-08-23 tooling bookkeeping review high fixed The claimed handover gate omitted promised checks and contradicted living artifacts
D-059 2026-08-23 tooling soundness flattering review critical outstanding The golden map mixed unhermetic discovery snapshots with mathematical pass-fail assertions
D-060 2026-08-23 tooling bookkeeping review high fixed Strict mode skipped deep regeneration and the atlas negative path was never exercised
D-061 2026-08-23 record validity flattering review high outstanding Unrecognised endpoint rows discard the evidence needed to classify them
D-062 2026-08-23 record soundness conservative inspection high fixed The D-042 correction left an executable n=12 rejection in the unattended runner
D-063 2026-08-23 docs soundness neutral inspection medium fixed A non-converse was called the contrapositive of the rigidity premise
D-064 2026-08-23 tooling robustness gate medium fixed Runner preflight was unable to run inside the gate that mutation-tested it
D-065 2026-08-23 record bookkeeping inspection low fixed The README repeated a numeric gate claim after declaring those counts removed
D-066 2026-08-23 docs soundness conservative inspection medium fixed The active baseline script still classified open n=12 as a negative control
D-067 2026-08-23 record bookkeeping review medium fixed The eleventh terminal round was omitted from campaign wall-time accounting
D-068 2026-08-23 record validity neutral review medium fixed The ledger labeled elapsed wall time as CPU time
D-069 2026-08-23 record bookkeeping inspection medium fixed H-002 still said its measured quench instrument was not built
D-070 2026-08-23 record bookkeeping inspection high fixed Exp-011 execution provenance was rewritten to a later record commit
D-071 2026-08-23 record bookkeeping inspection medium outstanding Generated session reports overwrite history and are not durable
D-072 2026-08-23 tooling robustness review high fixed Direct execute and release paths bypassed the gate marker
D-073 2026-08-23 record bookkeeping review medium fixed Agent-session filename and id agreement was not checked
D-074 2026-08-23 tooling bookkeeping review medium fixed The provenance regression tested receipt parsing but not artifact-field mapping
D-075 2026-08-23 docs soundness flattering review high fixed The PR 16 response claimed a portable oracle from an aggregate byte mismatch
D-076 2026-08-23 docs soundness neutral review high fixed Six distinct n=5 endpoint rows were treated as proof that identity was too fine
D-077 2026-08-23 record bookkeeping review high fixed The PR 16 handoff presented stale state and priorities as a current dependency queue
D-078 2026-08-23 docs soundness flattering review medium fixed The n=5 rank correction still named equality independence as the sufficient condition
D-079 2026-08-23 docs soundness flattering review medium fixed The handoff retained an invalid implication from record rigidity to non-record behavior
D-080 2026-08-24 docs validity flattering review high fixed The original neighbor-transfer criterion was satisfied by the cold-start grid
D-081 2026-08-24 tooling validity flattering review high outstanding Runner preflight treats any nonempty queue as enough unattended work
D-082 2026-08-24 docs validity conservative review high fixed The synopsis again generalized one n=17 regime to oblique-search blindness
D-083 2026-08-24 docs soundness flattering review high fixed A finite-quench residual was called evidence of attraction through epsilon 0.1
D-084 2026-08-24 docs soundness flattering review high fixed The n=11 frontier artifact asserted rigidity and the wrong gap rank
D-085 2026-08-24 tooling bookkeeping inspection medium fixed Living run commands could rewrite the locked dependency state
D-086 2026-08-24 record bookkeeping review high fixed The overnight agenda and current handoff retained obsolete work and effort state
D-087 2026-08-24 docs validity flattering review high fixed Angle-class evidence and an untested kink sweep combined distinct claims
D-088 2026-08-24 docs soundness conservative review high fixed The finite-n frontier missed newer certified n=68 and n=69 constructions
D-089 2026-08-24 docs soundness neutral review high fixed The n=17 record was given symmetric plus-or-minus 40-degree orientations
D-090 2026-08-24 docs soundness flattering review high fixed Two low-angle examples were generalized into a universal tiny-angle prior
D-091 2026-08-24 docs validity flattering review high fixed H-010 did not reproduce Stromquist's conditional proof mechanism
D-092 2026-08-24 docs soundness flattering review medium fixed Structured asymptotic data repeated the phantom Roth-Vaughan constant
D-093 2026-08-24 tooling soundness flattering review critical fixed Contact canonicalization discarded every advertised node attribute
D-094 2026-08-24 docs soundness flattering drift_check medium fixed The idea board repeated the wrong n=11 gap rank
D-095 2026-08-24 record bookkeeping drift_check medium fixed The idea board called a refuted hypothesis the best runnable experiment
D-096 2026-08-24 docs soundness conservative review high fixed Algebraic degree was promoted into a proof-method ceiling and obstruction
D-097 2026-08-24 docs soundness flattering review medium fixed Strategy catalogues were described as exhaustive histories of the field
D-098 2026-08-24 engine soundness flattering review medium fixed Source code still called a finite stochastic return threshold a basin radius
D-099 2026-08-24 docs validity flattering review high fixed H-023 placed six non-optimal endpoints in one optimum-side family question
D-100 2026-08-24 docs validity flattering review high fixed Delta continuation had no preserved parameter and reversed its progress metric
D-101 2026-08-24 record bookkeeping review high outstanding Historical quench round wall times disagree with retained per-call durations
D-102 2026-08-24 docs validity flattering review high fixed H-006 treated dual pose support as candidate piercing-point loci
D-103 2026-08-24 record validity flattering review high fixed H-011 omitted its declared component-identifiability prerequisites
D-104 2026-08-24 docs soundness flattering review high fixed Living artifacts still promoted local refinement failure into a wrong-basin claim
D-105 2026-08-24 docs validity flattering review medium fixed Superdisk continuation lacked a shape normalization and branch rules
D-106 2026-08-24 tooling bookkeeping gate medium fixed A living negative-control anchor still targeted corrected hypothesis prose
D-107 2026-08-24 docs bookkeeping gate high fixed Generated campaign views and next-work state lagged their corrected sources
D-108 2026-08-24 docs soundness flattering review high fixed A direct piercing-number application was missing from the method history
D-109 2026-08-24 docs soundness flattering review critical fixed Boundary corner coordinates were double-counted as independent wall equations
D-110 2026-08-24 docs validity flattering review high fixed A fixed-cell LP dual was mislabeled a free-framework self-stress
D-111 2026-08-24 docs validity flattering review high fixed A fixed-budget extreme-value fit was generalized to unlimited search support
D-112 2026-08-24 docs validity flattering review high fixed One fixed-cell angle sheet was presented as the global two-class landscape
D-113 2026-08-24 docs soundness flattering review high fixed Numerical branch coalescence was called a feasible-component merge and barrier
D-114 2026-08-24 docs validity flattering review high fixed Fractional piercing was given an integral and symmetric conclusion
D-115 2026-08-24 docs soundness flattering review medium fixed Negative novelty claims exceeded the searched literature surface
D-116 2026-08-24 docs bookkeeping review medium fixed The incoming strategy review mixed stale corpus state with current proposals
D-117 2026-08-24 docs soundness flattering review high fixed The idea board again said a basin-width question had been answered
D-118 2026-08-24 docs validity flattering review high fixed Budget scaling and attraction rarity were treated as the same cheap question
D-119 2026-08-24 record validity conservative review high fixed A continuous angle sheet was required to have only one nearby grid point
D-120 2026-08-24 quench validity neutral review high fixed Pair half-extent was recomputed four ways and injected ulp noise into cell selection
D-121 2026-08-24 tooling robustness review high fixed Invalid gate worker counts were accepted and a negative count could hang the throttle
D-122 2026-08-24 tooling validity flattering review high fixed A targeted gate built unrelated Rust code and could hide its build skip
D-123 2026-08-24 docs bookkeeping review medium fixed A per-step process cap was documented as a shared global worker budget
D-124 2026-08-24 tooling performance review medium fixed The portable negative-control fallback copied a needlessly broad repository surface
D-125 2026-08-24 tooling robustness gate high fixed Parallel negative controls could share one worker tree and erase each other's mutation
D-126 2026-08-24 quench validity neutral review high outstanding Quench convergence work was budgeted by machine time rather than reproducible work
D-127 2026-08-24 docs bookkeeping review medium fixed The engineering review called its branch complete before its delta blockers were resolved
D-128 2026-08-24 tooling performance review high fixed The packing gate serialized independent checks and offered no targeted edit loop
D-129 2026-08-24 tooling robustness review medium fixed Negative-control checker processes have no bounded timeout or child cleanup
D-130 2026-08-24 tooling bookkeeping gate medium fixed A defect-link mutation control reused an anchor that became nonunique
D-131 2026-08-24 tooling bookkeeping review medium fixed The Python lint floor passed while reporting eight type-checker warnings
D-132 2026-08-24 quench validity flattering review high fixed Fixed-cell iteration returns capped states without a settlement result
D-133 2026-08-24 record validity neutral pre_registered_rule high fixed Determination results could encode only search-specific basin outcomes
D-134 2026-08-24 docs bookkeeping inspection medium fixed The n = 29 frontier row omitted its analytic improver and verified angles
D-135 2026-08-24 docs bookkeeping inspection low fixed The experiment roll-up forbade the record-source evidence it displayed
D-136 2026-08-24 record soundness flattering review high fixed A branch linearization was mislabeled as the true Bouligand tangent
D-137 2026-08-24 tooling validity flattering review high fixed Trump certificate replay did not require one-to-one branch coverage
D-138 2026-08-24 tooling validity flattering anomaly high fixed The provenance gate skipped schema-valid unquoted commit hashes
D-139 2026-08-24 docs soundness flattering review high contained H-032 omitted directly relevant hard-square configuration-space literature
D-140 2026-08-24 docs soundness flattering control_cell high fixed The closed n = 3 sliding family was said to have one contact certificate
D-141 2026-08-24 tooling robustness anomaly medium fixed Byte-exact raw literature failed the generic Git whitespace check
D-142 2026-08-24 tooling validity conservative review high fixed The n = 3 source replay compared an unlabelled theorem with labelled homology
D-143 2026-08-24 tooling soundness flattering review high fixed The n = 4 source replay attributed an unreported f-vector to Alpert et al.
D-144 2026-08-24 tooling validity flattering review high fixed The orientation identity check compared two identical hard-coded tuples
D-145 2026-08-24 record bookkeeping review medium fixed A broad integration edit fixed the wrong defect record
D-146 2026-08-24 docs validity flattering inspection high fixed The Stromquist Figure 13 transcription changed its defining point set
D-147 2026-08-24 docs soundness flattering review high fixed H-010 substituted unit squares for Stromquist's strict open boxes
D-148 2026-08-24 docs validity flattering review high fixed The D-091 correction omitted the published Figure 14 unavoidability claim
D-149 2026-08-24 docs validity flattering review high fixed Search saturation was still presented as a Stromquist known-answer test
D-150 2026-08-24 record bookkeeping drift_check medium fixed Active effort and priority views drifted after four exact research rounds
D-151 2026-08-24 docs soundness flattering inspection high fixed Stromquist's middle Lemma 4 table selected an extraneous cubic root
D-152 2026-08-24 docs soundness flattering inspection critical fixed Stromquist's printed Figure 14 point set admits an escaping box
D-153 2026-08-24 docs bookkeeping inspection high fixed The source ledger called three directly hosted 1984 Stromquist memoranda unavailable
D-154 2026-08-24 verifier soundness flattering review high fixed The first H-010 checker draft treated unchecked polynomials as exact fields
D-155 2026-08-24 record robustness review medium fixed The first H-010 record draft retained raw platform floating-point diagnostics
D-156 2026-08-24 verifier soundness flattering review high fixed The first H-041 checker draft did not prove that tiling vertices stayed inside the container
D-157 2026-08-24 verifier soundness flattering review high fixed The first H-041 Lemma 6 replay squared inequalities without executable sign guards
D-158 2026-08-24 record validity flattering review medium fixed The first H-041 record blurred runtime source checks with unattached extraction hashes
D-159 2026-08-24 tooling robustness drift_check medium fixed Scanned archive PDFs were misclassified as text by Git
D-160 2026-08-24 record bookkeeping drift_check medium fixed A broad H-010 integration patch updated D-002 instead of D-151
D-161 2026-08-24 docs bookkeeping conservative drift_check medium fixed The synopsis did not track the live hypothesis count
D-162 2026-08-24 record validity flattering control_cell high outstanding The small-n golden convergence labels hid unsettled fixed-cell evaluations
D-163 2026-08-24 tooling validity flattering gate critical fixed The historical-regression gate masked a failing checker
D-164 2026-08-24 quench validity conservative control_cell high fixed The fixed-cell solver mislabeled a numerical rejection as mathematical infeasibility
D-165 2026-08-24 quench validity flattering control_cell critical fixed Initial cell-solve failures bypassed the fixed-point termination guard
D-166 2026-08-24 record soundness flattering drift_check critical fixed Basin events called producer convergence a fixed-cell certificate
D-167 2026-08-24 record bookkeeping drift_check high fixed Basin events omitted the wall time needed to price the loop
D-168 2026-08-24 quench validity conservative control_cell high fixed A finite equal-objective adjacent-cell closure was labeled an unsettled cycle
D-169 2026-08-24 quench validity flattering control_cell critical fixed The cell post-check ignored every containment row
D-170 2026-08-24 record bookkeeping neutral drift_check high fixed D-165 reused the unrelated D-132 bead
D-171 2026-08-24 quench robustness conservative control_cell high fixed Argmax-only n=4 repair left one positive-control evaluation unsettled
D-172 2026-08-24 docs validity flattering review high fixed The orientation tutorial said a configuration-space cell fixes the angles
D-173 2026-08-24 docs soundness conservative review high fixed The orientation docs ruled out equality certification by all interval methods
D-174 2026-08-24 docs soundness flattering review high fixed The recovery narrative silently eliminated centres and universalized rational LPs
D-175 2026-08-24 docs bookkeeping conservative drift_check high fixed The stacked orientation branch lagged the current basin calibration and agenda
D-176 2026-08-24 docs soundness flattering review high fixed The orientation tutorial promoted special-case angle evidence into general structure
D-177 2026-08-24 docs performance neutral review medium fixed The orientation docs asserted unmeasured continuity and language-independent cost
D-178 2026-08-24 docs bookkeeping flattering review medium fixed The orientation tutorial left project-specific theorem claims unbound to evidence
D-179 2026-08-24 docs soundness flattering review high fixed The orientation tutorial treated undefined hardness and s(11) tractability as settled
D-180 2026-08-24 docs bookkeeping conservative review high fixed Current orientation links sent agents to an explicitly superseded handoff
D-181 2026-08-24 docs soundness flattering review high fixed The basin definition called a deterministic quench endpoint a non-point
D-182 2026-08-24 docs validity conservative review medium fixed The strategy tutorial excluded safe loss shaping and all repeated endpoint hits
D-183 2026-08-24 record robustness flattering control_cell high fixed The basin-event batch crashed before retaining an independently invalid endpoint
D-184 2026-08-24 tooling validity conservative inspection low fixed Overlapping synopsis excerpts were mistaken for a duplicate experiment row
D-185 2026-08-24 docs bookkeeping inspection medium fixed The synopsis carried stale defect-direction and gate-detector aggregates
D-186 2026-08-24 tooling soundness flattering review high fixed The first angle-sheet bound omitted slide motion from one projection
D-187 2026-08-24 tooling bookkeeping conservative gate medium fixed Two synopsis mutation controls retained the pre-D-186 aggregates
D-188 2026-08-24 docs validity flattering review high fixed The conventions authorized multi-hypothesis rounds after the schema prohibited them
D-189 2026-08-24 docs validity flattering review high fixed The newcomer glossary omitted the point-basin and component boundary
D-190 2026-08-24 docs bookkeeping conservative review medium fixed The id glossary said every experiment retains a JSONL archive
D-191 2026-08-24 docs validity flattering review high fixed Orientation prose overstated both control automation and conventions ownership
D-192 2026-08-24 docs bookkeeping conservative review high fixed The stacked documentation PR was twelve commits behind the living synopsis
D-193 2026-08-24 docs bookkeeping flattering review medium fixed The draft PR 21 review misclassified a valid but incomplete definition as soundness
D-194 2026-08-24 tooling soundness flattering review high fixed The n=5 tangent checker reused one contact differential across three slide strata
D-195 2026-08-24 tooling soundness flattering review high fixed The n=5 tangent checker treated tied support inequalities as alternatives
D-196 2026-08-24 record bookkeeping flattering drift_check medium fixed A broad exp-035 status patch updated D-034 instead of D-194
D-197 2026-08-24 record bookkeeping neutral anomaly medium fixed A concurrent checkout moved the exp-036 instrument commit to the wrong branch
D-198 2026-08-24 tooling bookkeeping conservative gate medium fixed Two numeric negative-control anchors lagged the D-197 aggregates
D-199 2026-08-24 quench robustness conservative gate high fixed A frozen residual-repair set rejected a valid n=10 cell
D-200 2026-08-24 docs bookkeeping conservative inspection medium fixed The synopsis gate-defect enumeration omitted D-198
D-201 2026-08-24 tooling bookkeeping conservative review medium fixed Two mutation controls retained pre-integration expected diagnostics
D-202 2026-08-24 record bookkeeping conservative anomaly medium contained A delegated long-running validation lost its final receipt
D-203 2026-08-24 quench robustness conservative gate high fixed The n=4 seed-0 quench ends on a HiGHS solve error
D-204 2026-08-24 tooling bookkeeping conservative gate medium fixed The gate-aggregate control expected the mutated value
D-205 2026-08-24 record bookkeeping conservative review high fixed The stacked workflow PR retained a superseded scientific checkpoint
D-206 2026-08-24 record robustness flattering review high fixed The phase schema could not express the resumable contract promised by the docs
D-207 2026-08-24 docs validity neutral review medium fixed The workflow map handed repairs to an undefined implementation workflow
D-208 2026-08-24 docs validity neutral review medium fixed The new campaign definition excluded work already stored in the campaign
D-209 2026-08-24 docs bookkeeping conservative review medium fixed The legacy-series summary stopped at 35 of 36 experiments
D-210 2026-08-24 record validity flattering review high fixed Retrospective workflow labels were counted as contemporaneous declarations
D-211 2026-08-24 docs performance neutral review medium fixed The workflow cadence mandated review phases after routine guarded rounds
D-212 2026-08-24 docs validity neutral review medium fixed The workflow prose turned a primary focus into an exclusive focus
D-213 2026-08-24 tooling bookkeeping neutral review medium fixed Workflow vocabulary and entry state had several machine authorities
D-214 2026-08-24 tooling validity flattering review high fixed The v2 loader ignored its declared contract and envelope identity
D-215 2026-08-24 tooling bookkeeping flattering review medium fixed Workflow drift checks accepted unauthorized rows and invalid transitions
D-216 2026-08-24 tooling performance neutral review low fixed Full phase histories made the generated orientation table unboundedly wide
D-217 2026-08-24 record bookkeeping conservative anomaly medium contained A parallel validation wrapper discarded a live command receipt
D-218 2026-08-24 record bookkeeping conservative inspection medium fixed Session-009's output inventory stopped before exp-035 and exp-036
D-219 2026-08-24 record bookkeeping conservative review medium fixed The BC-010 agenda retained a superseded two-size solver diagnosis
D-220 2026-08-24 record bookkeeping flattering review high fixed The durable resume order was not encoded in the ready-work graph
D-221 2026-08-24 record bookkeeping neutral review medium fixed The campaign bead retained the superseded eight-hour horizon
D-222 2026-08-24 record robustness conservative anomaly medium contained A delegated read-only audit launched an explicitly excluded deep gate
D-223 2026-08-24 docs bookkeeping neutral review medium fixed The synopsis retained superseded normal-gate timing and control counts
D-224 2026-08-24 record validity conservative anomaly medium fixed A focused negative-control replay omitted the gate-managed Python environment
D-225 2026-08-24 record validity conservative review high fixed Checkpoint merge readiness was conflated with unattended launch readiness
D-226 2026-08-24 tooling robustness conservative gate medium fixed CI discarded the history required by its own provenance gate
D-227 2026-08-24 tooling performance neutral inspection low fixed The uv cache key prefixed its working directory twice
D-228 2026-08-24 tooling robustness conservative gate low fixed The workflow regression assumed every YAML key was a string
D-229 2026-08-24 record bookkeeping neutral review medium fixed The synopsis hypothesis-status aggregate drifted behind the generated ledger
D-230 2026-08-24 record validity flattering review high fixed Top-level synopsis claims exceeded the scope of the detailed lane evidence
D-231 2026-08-24 tooling bookkeeping neutral drift_check high fixed The synopsis checker omitted hypotheses with multiword statuses
D-232 2026-08-24 tooling bookkeeping neutral inspection medium fixed An unfrozen lint command rewrote the packing dependency lockfile
D-233 2026-08-24 tooling robustness conservative gate medium fixed README layout validation treated ignored migration caches as project content
D-234 2026-08-24 docs bookkeeping neutral review medium fixed The implemented engineering plan carried stale inventory and defect references
D-235 2026-08-24 docs bookkeeping neutral review high fixed Current reproducibility instructions still named deleted command paths
D-236 2026-08-24 record bookkeeping neutral review high fixed A completed timeout fix remained outstanding in the defect and launch records
D-237 2026-08-24 tooling robustness neutral inspection high fixed Mutation-control target paths could escape their private snapshot
D-238 2026-08-24 docs soundness flattering review high fixed The synopsis called quench endpoints discrete basins with a well-defined census
D-239 2026-08-24 tooling robustness neutral inspection medium outstanding Validation steps have no outer deadline or process-group cleanup
D-240 2026-08-24 tooling robustness neutral review high fixed The autonomous-session wall budget could be reset indefinitely
D-241 2026-08-24 record bookkeeping conservative review high fixed The session checker rejected a legitimate renewed slice
D-242 2026-08-24 tooling bookkeeping neutral inspection medium fixed A delegated focused check rewrote the dependency lock outside its scope
D-243 2026-08-24 docs bookkeeping flattering review high fixed The autonomous-loop instruction contradicted its upstream launch blocker
D-244 2026-08-24 record bookkeeping flattering review high fixed A completed delegation could shed provenance and pass an incomplete receipt
D-245 2026-08-24 record bookkeeping neutral review medium fixed A newly opened agent session had to invent completed progress
D-246 2026-08-24 tooling robustness neutral review medium fixed Reserve enforcement crashed on a malformed contemporaneous phase clock
D-247 2026-08-24 tooling bookkeeping neutral inspection medium fixed A delegated follow-up again rewrote the dependency lock outside its scope
D-248 2026-08-24 tooling robustness neutral review high fixed Active slice deadlines were recorded but not enforced after expiry
D-249 2026-08-25 record bookkeeping flattering inspection high fixed Branch completion and landed-main state were treated as the same fact
D-250 2026-08-25 tooling soundness flattering review high fixed A delegated exact checker asserted proofs and controls it had not executed
D-251 2026-08-25 tooling validity conservative control_cell medium fixed An omitted R6 coordinate produced a false exact blocker
D-252 2026-08-25 record bookkeeping conservative review medium fixed The preregistered refusal accidentally included already proved sheet lifts
D-253 2026-08-25 tooling robustness neutral gate medium fixed The exp-038 wrapper was added under a forbidden legacy tools path
D-254 2026-08-25 record soundness flattering review high fixed The first exp-039 criterion conflated ray normalizations and rank with dimension
D-255 2026-08-25 tooling bookkeeping conservative gate medium fixed Four exact negative-control anchors lagged the current living-document values
D-256 2026-08-25 tooling soundness flattering review high fixed Endpoint equality was treated as an affine full-interval axis proof
D-257 2026-08-25 tooling soundness flattering review high fixed Three stress samples were promoted without deriving the degree bound
D-258 2026-08-25 record bookkeeping flattering review high fixed An under-scoped status patch closed the wrong outstanding defect
D-259 2026-08-25 record bookkeeping conservative drift_check medium fixed An explicit Flowmark target bypassed the generated-view exclusion
D-260 2026-08-25 tooling validity flattering control_cell high fixed A solver fixture cross-wired two fixed-point evaluations
D-261 2026-08-25 tooling validity flattering review high fixed The replay fixture was allowed to define its own failure receipt
D-262 2026-08-25 tooling bookkeeping flattering gate medium fixed A delegated fixture stopped after pytest while static checks were red
D-263 2026-08-25 record bookkeeping flattering review high fixed A second status-only patch closed D-034 instead of the named defect
D-264 2026-08-25 record bookkeeping conservative gate medium fixed An active session ended its workflow list on a completed phase
D-265 2026-08-25 record bookkeeping flattering gate medium fixed Two delegation deadlines exceeded their declared budgets
D-266 2026-08-25 quench soundness conservative review high fixed A numerical fallback disagreement was promoted to infeasibility
D-267 2026-08-25 tooling bookkeeping conservative gate medium fixed The soundness mutation diagnostic kept the old denominator
D-268 2026-08-25 tooling bookkeeping conservative gate medium fixed Focused fallback checks omitted the Python format gate
D-269 2026-08-25 record bookkeeping conservative gate low fixed A session delegation used an undeclared timing-quality value
D-270 2026-08-25 record bookkeeping conservative review low fixed Phase 11 retained a next action that its own evidence had completed
D-271 2026-08-25 record bookkeeping conservative gate low fixed A phase rotation used an undeclared transition label
D-272 2026-08-25 tooling bookkeeping conservative gate medium fixed The macOS expected-failure guard rejected a recovered deep golden
D-273 2026-08-25 tooling validity flattering review high fixed The direct-blocking workflow test admitted advisory variants
D-274 2026-08-25 tooling bookkeeping conservative gate low fixed The delegated promotion check named deleted paths and left one long assertion
D-275 2026-08-25 engine validity flattering review high fixed The first pair-meter draft omitted basin-entry work from its summary
D-276 2026-08-25 tooling bookkeeping conservative gate low fixed The first basin-entry meter probe selected a fixture without angles
D-277 2026-08-25 tooling bookkeeping conservative gate low fixed The first pair-meter integration test omitted crate documentation
D-278 2026-08-25 record bookkeeping conservative gate low fixed The D-277 regression command made the defect source invalid YAML
D-279 2026-08-25 record bookkeeping conservative gate low fixed The finalization phase used an undeclared focus value
D-280 2026-08-25 tooling performance conservative gate high contained The phase-count cap ended a fast campaign before its wall-clock goal
D-281 2026-08-25 record bookkeeping conservative gate medium fixed New defect totals left synopsis controls and the open inventory stale
D-282 2026-08-25 record bookkeeping conservative gate low fixed The stopped session retained a completed final phase
D-283 2026-08-25 engine robustness neutral review high outstanding A zero-step sqsearch regime can spin forever without spending its move budget
D-284 2026-08-25 record validity flattering review medium fixed A delegated float scout ran before the Trump support criterion was frozen
D-285 2026-08-25 tooling validity flattering review high fixed The first H-043 pilot could pass while its exact cone oracle was unresolved
D-286 2026-08-25 record bookkeeping neutral review medium fixed The first H-043 record conflated conjunctive wall rows with feature aliases
D-287 2026-08-25 record bookkeeping neutral anomaly low fixed A shell-active bead description executed a backtick-delimited word
D-288 2026-08-25 tooling bookkeeping neutral gate low fixed The first focused Ruff replay doubled the uv working-directory prefix
D-289 2026-08-25 tooling validity neutral review high contained The H-043 pilot normalized row classes for counting but not for its cone oracle
D-290 2026-08-25 record robustness conservative review medium outstanding H-043 pilot certificates can be recomputed but not replayed from a retained record
D-291 2026-08-25 tooling validity conservative review high contained The branch-0 H-043 golden would classify a valid future refutation as test failure
D-292 2026-08-25 record validity neutral review high contained H-043 branch indices are regenerated but not bound to the retained exp-013 universe
D-293 2026-08-25 tooling validity flattering review medium outstanding Two H-043 provenance selftests claim completeness without checking exact mappings
D-294 2026-08-25 record bookkeeping neutral inspection medium fixed The canonical defect count lagged eleven committed records
D-295 2026-08-25 tooling robustness neutral review high fixed The first timeout draft isolated validation calls that remained unbounded
D-296 2026-08-25 tooling validity flattering review high fixed The first timeout test could not expose an early-parent descendant leak
D-297 2026-08-25 tooling robustness neutral review high fixed The first Windows timeout path treated taskkill as a proved tree guarantee
D-298 2026-08-25 tooling validity neutral review low fixed Overlapping inspection ranges fabricated a duplicate session key
D-299 2026-08-25 tooling bookkeeping flattering review medium fixed A later successful check masked an earlier gate failure in the shell status
D-300 2026-08-25 record bookkeeping conservative pre_registered_rule medium outstanding The receipt rehearsal requested an unsupported gdate precision
D-301 2026-08-25 docs bookkeeping neutral review low fixed The first receipt rule split a compound adjective after its hyphen
D-302 2026-08-25 docs soundness conservative review high fixed Bui transcription changed the square-reduction divisor into a product
D-303 2026-08-25 docs soundness flattering review high fixed Bui transcription omitted the quantified width hypothesis
D-304 2026-08-25 docs soundness flattering review high contained McClenagan's Section 3 proof uses a contradictory inequality chain
D-305 2026-08-25 docs bookkeeping flattering gate medium fixed The source checkpoint undercounted fixes without regression checks
D-306 2026-08-25 tooling bookkeeping neutral gate medium fixed Four aggregate mutation controls retained stale exact anchors
D-307 2026-08-25 tooling validity conservative gate medium fixed Host Python misdiagnosed valid Python 3.14 exception syntax
D-308 2026-08-25 tooling bookkeeping neutral control_cell medium fixed Two synchronized mutation controls expected the canonical rather than mutated count
D-309 2026-08-25 record bookkeeping neutral gate low fixed The first session-012 draft violated two clocked-session fields
D-310 2026-08-25 docs soundness flattering review medium fixed McClenagan's equation 3.2 calls a periodic tangent solution unique
D-311 2026-08-25 tooling validity flattering control_cell high fixed The open-defect bead control still mutated a fixed record
D-312 2026-08-25 record bookkeeping neutral gate low fixed An early checkpoint phase claimed the outer finalization clock role
D-313 2026-08-25 record bookkeeping conservative drift_check medium fixed Explicit Flowmark target bypassed the generated-ledger exclusion
D-314 2026-08-25 tooling robustness neutral review high fixed The production commit-state probe was initially unbounded
D-315 2026-08-25 tooling robustness neutral review high fixed Detached production defaults could still wait on interrupted workers
D-316 2026-08-25 tooling robustness neutral review high fixed Quiet provenance probing lacked a Windows fail-closed guard
D-317 2026-08-25 tooling validity conservative gate medium fixed The SIGINT timeout regression ceiling was too tight for the focused suite
D-318 2026-08-25 tooling bookkeeping conservative control_cell low fixed The unprotected-fix mutation control retained a stale expected count
D-319 2026-08-25 record bookkeeping conservative review high fixed The cold-start handoff pointed to a closed BC-010 bead
D-320 2026-08-25 tooling robustness conservative gate medium fixed Deep golden validation treated YAML line wrapping as scientific drift
D-321 2026-08-25 tooling validity conservative gate medium fixed The SIGINT cleanup regression signaled before its child start condition
D-322 2026-08-25 docs bookkeeping neutral review low fixed The synopsis overstated which defect classes the gate had caught
D-323 2026-08-25 docs soundness flattering review high fixed The reader-first frontier view discarded long exact forms
D-324 2026-08-25 tooling soundness flattering review medium fixed Semantic golden comparison accepted ambiguous YAML identities
D-325 2026-08-25 tooling robustness conservative review low fixed The current-handoff validator embedded one live agenda cell
D-326 2026-08-25 docs bookkeeping flattering review medium fixed A second unprotected-fix aggregate escaped the single-match drift check
D-327 2026-08-25 docs bookkeeping neutral review low fixed The timeout documentation split another compound adjective after its hyphen
D-328 2026-08-25 record bookkeeping neutral gate medium fixed Concurrent-branch rounds were authored to a superseded experiment schema
D-329 2026-08-25 record soundness neutral review medium fixed The tutorial's promotion example asserted an unattested separation value
D-330 2026-08-25 record soundness neutral review medium fixed The tutorial identified the side floor with the feasibility tolerance
D-331 2026-08-25 record soundness neutral review low fixed A benchmark ratio was relabeled across implementations
D-332 2026-08-25 record bookkeeping flattering review low fixed The tutorial overstated archive coverage of the record constructions
D-333 2026-08-25 record bookkeeping conservative review medium fixed The registry's rounds column drifted from the ledger under no stated rule
D-334 2026-08-25 record bookkeeping flattering review low fixed A hand-copied no-regression count drifted beside a sentence promising no drift
D-335 2026-08-25 docs bookkeeping neutral review medium fixed The tutorial's moduli figure went unexplained and its axis collided with the text
D-336 2026-08-25 docs bookkeeping flattering review medium fixed Verified claims carried no provenance axis, blurring confirmations and new results
D-337 2026-08-25 docs bookkeeping neutral review medium fixed An upstream merge restored stale synopsis totals and an incorrect H-024 disposition
D-338 2026-08-26 docs bookkeeping neutral gate medium fixed The atlas handoff bypassed the declared SVG exposition gallery
D-339 2026-08-26 tooling robustness neutral gate high fixed The retained chunk census depended on platform libm rounding
D-340 2026-08-26 tooling validity flattering review critical fixed The Bash gate suppressed errexit inside multi-command steps
D-341 2026-08-26 tooling validity neutral review high fixed In-progress experiments were forced to invent scientific results
D-342 2026-08-26 tooling robustness neutral review high fixed Offset leases were compared after discarding their UTC offset
D-343 2026-08-27 docs soundness neutral review medium contained Bui's replacement instruction omitted its finite upper index bound
D-344 2026-08-27 docs soundness flattering inspection high contained El Moumni's n=7 proof prints a negative segment length
D-345 2026-08-27 docs soundness flattering inspection high contained El Moumni's n=7 theorem drops Proposition 2's minimum branch
D-346 2026-08-27 docs soundness flattering inspection high contained El Moumni's Figure 4 chord line passes through the wrong square center
D-347 2026-08-27 docs soundness flattering inspection high contained El Moumni's Figure 4 diameter argument uses an undefined point i
D-348 2026-08-26 tooling performance neutral inspection medium fixed Raw quench traces expanded into thousands of timeline controls
D-349 2026-08-28 quench bookkeeping conservative review low contained An aborted free sweep drops its LP work from the run counter
D-350 2026-08-28 tooling bookkeeping neutral review medium fixed A rejected quench left the previous run's trace reachable from Download
D-351 2026-08-28 tooling bookkeeping flattering review medium fixed The replay command promised byte fidelity and compared only semantics
D-352 2026-08-28 tooling bookkeeping neutral review medium fixed A timeline counter changed meaning on the last event of every trace
D-353 2026-08-28 docs bookkeeping flattering review medium fixed The lab told users the optimizer receives a container side it never sees
D-354 2026-08-28 record soundness conservative review high fixed The frontier encoded catalogue silence about rigidity as a false rigid flag
D-355 2026-08-28 tooling performance neutral review high fixed Verification runs the whole gate after every change, so cycle time is set by the slowest step rather than by the change
D-356 2026-08-29 tooling validity conservative control_cell low outstanding The negative-control snapshot prunes the archive, so the n=29 chain cannot be controlled there
D-357 2026-08-29 tooling robustness control_cell low outstanding One synopsis negative control failed reproducibly, then passed, with no identified trigger
D-358 2026-08-29 record bookkeeping review medium fixed An unattended run misread its own clock by a factor of four and stopped early
D-359 2026-08-29 tooling robustness gate medium fixed The generated atlas SVG's coordinate precision is inherited from ambient state
D-360 2026-08-29 record validity neutral review low fixed A null-space finding was read off a filtered display instead of being computed
D-361 2026-08-29 tooling soundness conservative control_cell high fixed An edge-edge contact was assembled as one equation, so the contact system did not determine the pose
D-362 2026-08-29 tooling robustness review low outstanding The translation-only trajectory check compares independently rounded projections for exact equality
D-363 2026-08-29 tooling validity neutral review low fixed A closure form was named in the record that could not close the case it was named for
D-364 2026-08-29 tooling soundness flattering review low fixed The margin rule counted digits the value does not carry
D-365 2026-08-29 quench robustness gate medium outstanding The deep golden-basin oracle fails at n = 10, and only the strict tier runs it
D-366 2026-08-29 tooling performance gate low fixed The negative-control step outgrew the gate's per-step timeout
D-367 2026-08-29 record soundness flattering review medium fixed Degree bounds on the Kingbird solution were written as bounds on s(n)
D-368 2026-08-29 docs performance review medium fixed The rules an agent needs before its first tool call had nowhere to live
D-369 2026-08-29 tooling performance review medium fixed The gate's cheap checks were hostage to its expensive one, so drift reached CI
D-370 2026-08-29 tooling performance review medium fixed The registry checks were slow because they were not registry checks
D-371 2026-08-29 tooling robustness review low fixed The control snapshot guard was one artifact from refusing to run at all
D-372 2026-08-30 docs bookkeeping review medium fixed The reader-facing tier said four built capabilities were unbuilt, for four days
D-373 2026-08-30 record validity flattering control_cell medium fixed The identity acceptance rule named only controls whose answer was one
D-374 2026-08-30 record validity flattering inspection medium fixed The agenda queue offered four finished commitments as takeable
D-375 2026-08-30 record validity flattering control_cell medium fixed The atlas relation was scored at the wrong level, and the n=4 control separates nothing
D-376 2026-08-30 record bookkeeping neutral inspection low fixed The n=5 pair D-034 has quoted since August was never retained
D-377 2026-08-30 record bookkeeping neutral inspection low contained A read-only delegation cannot be recorded while it is running
D-378 2026-08-30 record validity flattering review medium contained No retained control can separate the surviving identity relation from merging everything
D-379 2026-08-30 record bookkeeping flattering review low fixed An operating rule understated its own measured cost by quoting the transcript, not the rollup
D-380 2026-08-30 tooling robustness conservative anomaly low fixed A superseded CI run reported the required check as a hard failure
D-381 2026-08-30 tooling validity conservative gate low fixed The pre-push tier is blind to a test that pins a string in the code it tests
D-382 2026-08-30 record bookkeeping flattering review low fixed A sentence asserted the rigidity its own next paragraph withdrew
D-383 2026-08-30 tooling robustness neutral anomaly medium fixed A cleanup error in the gate's own lock discarded a completed run's entire report
D-384 2026-08-30 tooling bookkeeping neutral control_cell low fixed A count keyed on id() moved between two identical runs
D-385 2026-08-30 record bookkeeping flattering review medium fixed The composite figure badges rigidity from a hard-coded set and never reads the frontier
D-386 2026-08-30 record bookkeeping neutral gate low fixed A session declared two phases starting an hour and a half after the clock read
D-387 2026-08-30 record bookkeeping neutral review low fixed A rule survived in one carried reminder and was dropped by the next
D-388 2026-08-30 tooling soundness flattering control_cell high fixed A rationalization that was exhaustive at one size answered the wrong system at another
D-389 2026-08-30 record validity conservative review medium fixed A route was priced without asking whether its destination was already reachable
D-390 2026-08-30 tooling soundness flattering control_cell high fixed An incidence was read as a contact, and the difference is a false constraint
D-391 2026-08-30 tooling soundness flattering control_cell high outstanding A tangent cone that is a union of half-spaces was intersected
D-392 2026-08-30 tooling validity flattering gate medium fixed A size cutoff meant for generated blobs went blind on a source file that grew
D-393 2026-08-30 tooling bookkeeping neutral gate medium contained The pre-push floor does not run tests, and CI stayed red for seventy-five minutes
D-394 2026-08-30 tooling bookkeeping conservative gate low fixed A contract sweep counted its own guard's filename as a use of the field
D-395 2026-08-30 record robustness review high fixed An unattended run stopped on a budget it had set for itself
D-396 2026-08-30 record validity flattering review medium fixed A case's rigidity block could claim verified over numerical evidence
D-397 2026-08-30 record validity conservative review medium fixed Two more sub-agents reported valid Python 3.14 as a syntax error
D-398 2026-08-30 record bookkeeping review medium fixed Three cases declare a mathematical blocker the repository has already cleared
D-399 2026-08-30 docs bookkeeping inspection low fixed A residual's explanation had the sign backwards, and called a rounding a truncation
D-400 2026-08-30 record bookkeeping review medium fixed A blocker denied a certificate the register held, and two guards let it
D-401 2026-08-30 record bookkeeping review low fixed The queue guarded against over-claiming readiness and not against hiding it
D-402 2026-08-30 record validity flattering gate high fixed A contact structure was read out of padding and reported as a result
D-403 2026-08-30 tooling robustness review medium contained The negative controls do not run on a pull request, so stale ones accumulate unseen
D-404 2026-08-31 record bookkeeping neutral review low fixed exp-045's registration froze a mutation vocabulary its instrument never carried
D-405 2026-08-31 record validity conservative review medium contained A sequencing plan was built on an unmeasured claim that the verified gap ranks nothing
D-406 2026-08-31 record bookkeeping review medium outstanding BC-092 was stopped on an enumeration price that traces to no artifact
D-407 2026-08-31 tooling validity conservative gate medium fixed The control snapshot omitted every file the checked documents link to
D-408 2026-08-31 tooling validity conservative gate medium fixed A negative control was defused by its collision partner completing
D-409 2026-08-31 tooling validity conservative gate medium fixed The pruned-link copy covered the archive but not the pruned renderings
D-410 2026-08-31 tooling validity conservative gate low fixed A control's anchor was bumped without its expected message
D-411 2026-08-31 tooling validity neutral review high fixed The results checker and its epistemic policy disagreed at the boundary rungs
D-412 2026-08-31 tooling validity flattering review high fixed Change-scoped validation could skip the open-ended results register dependencies
D-413 2026-08-31 tooling validity conservative review medium fixed Results controls ran in snapshots missing their structured evidence files
D-414 2026-08-31 tooling robustness review medium fixed CI installed an unpinned uv release
D-415 2026-08-31 tooling validity conservative gate low fixed Flowmark reflow defused the README defect-aggregate control anchor
D-416 2026-08-31 record validity flattering review high fixed The results scope contract admitted a broader claim and reversed ranges
D-417 2026-08-31 tooling validity conservative gate medium fixed The n=5 identity-pair replay compared meaningless floating-point last bits
D-418 2026-09-01 tooling validity flattering review high outstanding The BC-140 bound checker passed eight bounds with no named exceeding mutation
D-419 2026-09-01 record validity flattering review high outstanding Exp-057 applied six-decimal coordinate semantics to a fourteen-digit side token
D-420 2026-09-01 tooling validity flattering review high outstanding The BC-142 reachable-tests control did not prove exact selection equivalence
D-421 2026-09-01 tooling robustness gate medium outstanding The wave-efficiency renderer refused the mixed Claude-to-Codex agenda lanes
D-422 2026-09-03 tooling robustness inspection high fixed Build caches are still copied into every worker, and the cap is held by a per-file prune
D-423 2026-09-03 record bookkeeping review medium outstanding The cover certificate's Lemma 4 boundary is non-strict in two sources and strict in a third, unrecorded
D-424 2026-09-03 tooling validity flattering review high fixed Two of the eight n = 5 rigidity controls adjudicated themselves and called nothing
D-425 2026-09-03 tooling robustness inspection medium fixed A commit titled "restore" shipped the widest disabled runner guard to a shared branch
D-426 2026-09-03 record validity flattering review high outstanding A control-cell breach does not stop the unattended session
D-427 2026-09-03 tooling validity flattering inspection medium outstanding Controls are scored on the mutated run alone, so one can pass over an already-red checker
D-428 2026-09-03 verifier validity flattering review medium outstanding The n = 17 successor validator never ties the rebuilt chain spine to the carried boundary
D-429 2026-09-03 tooling robustness inspection low outstanding The generated-view comparison folds an ellipsis but not the space the formatter puts before it
D-430 2026-09-04 docs bookkeeping review low fixed The frontier prose kept saying sixty-three open cases rest on Nagamochi after the adoption made it sixty
D-431 2026-09-04 record bookkeeping review medium outstanding T-009's significance rationale compares its interval certificate against a different packing
D-432 2026-09-04 tooling robustness gate medium fixed The push tier keeps the 900-second cap when its selector picks the whole suite
D-433 2026-09-04 tooling robustness review medium fixed The certificate rationaliser floored its weights while its docstring said it rounded up
D-434 2026-09-04 tooling validity conservative inspection high fixed The covering LP's separation oracle sees fewer placements than the verifier decides
D-435 2026-09-04 verifier soundness flattering inspection high fixed An enclosing interval run accepted a certificate whose least mass it had pinned below one
D-436 2026-09-04 verifier bookkeeping review medium fixed The third-party package misnamed its own control and refused malformed files by traceback
D-437 2026-09-04 tooling robustness gate high fixed A strong-duality test asserted float equality and failed on half of CI runs
D-438 2026-09-04 tooling performance gate high fixed The fast test tier outgrew its budget as the certificates grew, and hid a failure
D-439 2026-09-04 docs bookkeeping inspection medium fixed Three records described the top rung, and the top rung moved out from under them
D-440 2026-09-04 record bookkeeping inspection medium fixed A confirmation rung rested on a decision no control ever made
D-441 2026-09-04 tooling robustness inspection medium fixed The generator decided its candidate before writing it, so a kill lost the candidate
D-442 2026-09-04 record bookkeeping inspection medium fixed A case body kept quoting a bound its own front matter had already moved past
D-443 2026-09-04 record bookkeeping review low fixed A generated document counted stop reasons its own sources do not record
D-444 2026-09-04 verifier bookkeeping gate medium fixed Three tests pinned the Nagamochi count, and the poisoned control went blind a second time
D-445 2026-09-04 record bookkeeping review medium fixed The n = 11 case body stayed on the 19/5 rung under front matter that said 381/100
D-446 2026-09-04 record bookkeeping review low fixed Three records said C5 needs a reviewer outside the project; the rubric says review-ready
D-447 2026-09-04 record soundness conservative review low fixed The certificate ceiling was written up as a method-wide impossibility
D-448 2026-09-04 record bookkeeping review low fixed Two comparison factors in T-017's prose were wrong by four times
D-449 2026-09-04 verifier validity neutral review low fixed The exact sweep's reported witness centre lies outside the admissible domain on most directions
D-450 2026-09-05 record bookkeeping review high fixed The introducing commit named the wrong SHA-256 for the retained 381/100 certificate
D-451 2026-09-05 record bookkeeping review medium fixed Three n = 12 records described rungs the retained pointer had already climbed past
D-452 2026-09-05 docs bookkeeping inspection medium fixed The documents a reader opens first still described the pre-T-018 n = 11 rung
D-453 2026-09-05 record soundness flattering review low fixed Every open case body rounded its bound to nearest, and so stated one stronger than is proved
D-454 2026-09-05 record bookkeeping review medium fixed Four durable records described a rung, a ladder and a count that had already moved
D-455 2026-09-05 tooling robustness gate medium fixed The inline-SVG ownership check swept the vendored submodule's own fixtures
D-456 2026-09-05 tooling performance gate medium fixed The exhaustive exact tier outgrew its budget, and the ceiling cited against raising it never existed
D-457 2026-09-05 tooling bookkeeping inspection low fixed The atlas drew one line weight for cairosvg and another for a browser
D-458 2026-09-05 record bookkeeping gate medium fixed A frozen-artifact path named a moving pointer, so a superseded rung quoted its successor's mass
D-459 2026-09-05 tooling robustness inspection low fixed A conflicted branch produced no CI at all, and no checks reads as pending rather than red
D-460 2026-09-05 record bookkeeping gate low contained A research cell with no hypothesis cannot produce a round, and BC-206 was written that way
D-461 2026-09-05 tooling bookkeeping inspection low fixed The Pages filter named the composite PNG and PDF but not the SVG the figure shows
D-462 2026-09-05 tooling bookkeeping inspection low fixed A plain git clone fails the documentation check
D-463 2026-09-05 tooling bookkeeping inspection low fixed Three canvases took every touch, so the page could not be scrolled past them
D-464 2026-09-05 tooling bookkeeping inspection low fixed Every link on the page failed contrast in dark mode
D-465 2026-09-05 tooling bookkeeping gate low fixed The gate counted its own render output against the mutation-snapshot cap
D-466 2026-09-05 tooling performance inspection medium fixed The pull-request tier tripled inside a cap with 3.61x of headroom, and no check could have objected
D-467 2026-09-05 record bookkeeping inspection medium outstanding Two branches allocated the same three defect ids to different defects, and nothing local can see it
D-468 2026-09-05 tooling validity neutral anomaly high fixed The record gate judged four refusals against the wall clock, so a green commit did not stay green
D-469 2026-09-06 tooling bookkeeping inspection low fixed A dependency comment claimed a reproducibility the code it describes denies
D-470 2026-09-06 tooling bookkeeping gate medium contained A test pinned to the moving certificate pointer went stale, and no pull request could have caught it
D-471 2026-09-06 tooling robustness inspection medium contained The conflicted-branch CI blackout fired three more times in one day, and nothing but a person detects it
D-472 2026-09-06 record bookkeeping gate medium outstanding A single CI timing was treated as a measurement, and a twenty per cent difference was attributed inside thirty-four per cent of noise
D-473 2026-09-06 verifier soundness flattering review high fixed Ceiling verification trusted a floating screen outside its justified numerical range
D-474 2026-09-06 record validity conservative review medium fixed Feasible covering masses were subtracted to infer a lower slope and a stopping side
D-475 2026-09-06 docs validity conservative review medium fixed The README presented a fixed-net numerical ceiling without its net restriction
D-476 2026-09-06 verifier soundness flattering review high fixed Cutting separation could retain an invalid scaled bound before final verification
D-477 2026-09-06 verifier soundness flattering review high fixed Cutting separation merged distinct exact vertices with identical float coordinates
D-478 2026-09-06 record validity flattering review medium contained The BC-206 cutting floor has no retained family for independent replay
D-479 2026-09-07 docs soundness flattering review medium fixed Case summaries replaced cited geometric helper proofs with generic pure dots counting
D-480 2026-09-07 tooling validity conservative inspection low fixed Raised footnote ink shifted the list-marker alignment measurement
D-481 2026-09-07 record bookkeeping conservative review medium fixed The n26–27 source fields omitted Green's stronger reported lower bound
D-482 2026-09-07 record bookkeeping conservative review medium fixed Expanded frontier source fields omitted stronger indexed lower-bound reports
D-483 2026-09-07 record bookkeeping conservative review medium fixed Case prose accepted a stale exact verified bound as a rounded current value
D-484 2026-09-08 tooling robustness conservative gate high fixed The slow behavioural lane ran serially and the escape screen shared its runner
D-485 2026-09-08 tooling robustness anomaly high fixed A glyph-centering override stretched CSS bullet squares into vertical bars
D-486 2026-09-08 tooling robustness review medium fixed Font pruning ignored the saved sans serif reading preference
D-487 2026-09-08 tooling validity flattering review medium fixed Detaching a font inspection session silently ended print emulation
D-488 2026-09-08 tooling robustness neutral gate high fixed The pre-push tier ran its whole-suite step in one process and could not finish it
D-489 2026-09-09 tooling validity flattering review high outstanding The corner-dual salvage screen can only accept a family that already fails the mass threshold
D-490 2026-09-10 tooling robustness gate medium contained Two renders of the explainer disagreed by two bytes and the check could not say where
D-491 2026-09-11 tooling validity conservative gate medium fixed The math geometry probe mixed observations and left a page watchdog uncontrolled
D-492 2026-09-19 tooling validity conservative review high fixed Threshold producer called solver errors infeasible
D-493 2026-09-19 verifier soundness conservative review high fixed Event-cell geometry accepted directions outside its quadrant
D-494 2026-09-19 tooling robustness review medium fixed Covering queue had no atomic writer ownership
D-495 2026-09-19 record validity conservative review high fixed Unfinished covering and interval checks were treated as refutations
D-496 2026-09-19 record bookkeeping review medium fixed Floating solver subjects were labelled exact and verified
D-497 2026-09-19 record bookkeeping review medium fixed T-027 provenance was swallowed by a YAML prose scalar
D-498 2026-09-19 record bookkeeping review medium fixed Survey census omitted the three proved cases at 98 through 100
D-499 2026-09-19 record bookkeeping review medium fixed Doubled certificate direction count was copied incorrectly
D-500 2026-09-19 tooling soundness flattering review medium fixed Threshold columns silently discarded site multiplicities
D-501 2026-09-19 record bookkeeping gate medium fixed Retained atlas graphics lagged behind the T-030 bound
D-502 2026-09-19 tooling robustness gate high outstanding Costly partial pre-push selections kept pytest serial until the command timed out
D-503 2026-09-19 tooling robustness gate medium fixed Artifact timeout test assumed a Python child could print within 200 milliseconds
D-504 2026-09-19 tooling performance gate medium fixed Allowlist existence test generated a full declared-bound reference report
D-505 2026-09-20 docs soundness conservative review high fixed Bentz 2016 transcription moved the finishing line of Theorems 9 and 11
D-506 2026-09-20 docs soundness conservative review high fixed Bentz 2016 transcription inverted Lemma 7's bound and rewrote its lemma reference
D-507 2026-09-20 docs soundness conservative review high fixed Bentz 2016 transcription dropped the factor 2 from Theorem 9's budget line
D-508 2026-09-24 record bookkeeping flattering inspection low fixed Rung-0 records counted the reader's branch nodes as leaf certificates
D-509 2026-09-25 tooling robustness gate medium contained One math glyph's baseline is decided per page load, so drawing each load twice does not contain it
D-510 2026-10-01 verifier soundness flattering review high fixed Independent rational checker ignored the declared coordinate frame
D-511 2026-10-01 tooling robustness review medium fixed Half-angle witness export declared a schema absent beside its output
D-512 2026-10-01 tooling bookkeeping conservative drift_check medium fixed The deployed site dropped every record link into the archive and the campaign
D-513 2026-10-02 tooling bookkeeping conservative gate medium fixed The site's table-layout tests skipped on every pull request, and read hinted text on Linux
D-514 2026-10-02 docs soundness flattering inspection medium fixed The Erdős–Graham transcription stated Theorem (1) as a two-sided bound the paper does not prove
D-515 2026-10-02 docs soundness flattering inspection medium fixed The McClenagan transcription swapped Montgomery's exponent, and the record then faulted McClenagan for it
D-516 2026-10-02 record soundness flattering review high fixed The register held Nagamochi 2005's Theorem 2 as verified and rested 287 case floors on it, though its Lemma 1 is false
D-517 2026-10-04 tooling soundness flattering review low fixed Borrowed lower bounds were rounded to nearest, and the checker's two-sided tolerance passed a floor rounded up
D-518 2026-10-04 tooling validity flattering review medium fixed The Lemma 1 checker's Karakuş ordering tested only k^2 <= N, and the evidence register called it an exact check
D-519 2026-10-05 record bookkeeping conservative inspection medium fixed Three Kingbird counts held as pending catalogue intake named no owning bead, and the record trailed the catalogue for five days
D-520 2026-10-12 tooling bookkeeping conservative inspection medium fixed The open-gap ranking filtered on the reported status, so a case reported closed but open in the verified lane left the table
D-521 2026-10-13 verifier soundness flattering review medium outstanding Numerical witness receipts record tolerance-blurred values as the minimum clearance and pair gap
D-522 2026-10-15 record bookkeeping conservative inspection medium fixed Beyond the corpus the record compared reports only with the grid and earlier reports, so two current rows stood above Göbel's classical constructions
D-523 2026-11-12 tooling robustness gate medium fixed The site layout step's stack-dump timer crashed an xdist worker when a preview-server request thread ended during the dump