The answer is "it depends what you call an information set", and the
distinction is the result. 44,938 decision points, random play, 2/3/4/6
seats.
Reading A — information set = the seat's current projection, which is what
project(Viewer::Player(seat)) returns and what the page renders: 22
violations. Reading B — information set = the seat's observation history,
every view seen and action taken in order: 0.
The Reading A witness is concrete. Two histories reach a byte-identical
view — round 3, Select step, same hand, same claimed Problem — where the
seat had played SOLVE then GROUND-OU(protect) in one and SUPPORT then
SOLVE in the other. The view does not tell the seat what it did, because
our state is a snapshot rather than a history: selections clear each round
and effects coincide, so a player cannot reconstruct their own past from
the present. In a real game the player's memory supplies it; in the state,
nothing does.
That is precisely OpenSpiel's ObservationString vs InformationStateString
split, arrived at here by measurement rather than read off. project() is
an observation, not an information state.
So Track B is not closed, it is constrained, and usefully: an
extensive-form game built from this engine must key information sets on
observation histories, never on project().
Both directions are asserted — Reading B empty AND Reading A non-empty —
because if the sample stops finding Reading A violations the conclusion is
unsupported and must be re-derived rather than quietly kept. And the check
samples, so it can falsify perfect recall and cannot establish it: Reading
B's zero means no counterexample was drawn, which is printed as such.
Wired into make panels, so it is re-derived by the gate rather than by
hand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The pattern is now measured over three rounds: 5 fatal, then 3 (2 from the
previous round's corrections), then 4 (3 from them). The corrections are
not getting safer.
FATAL 1: round 2's short-cell assertion went into regulation.rs only.
attack-value.rs — which produced every number in CB-EV-0030's DARVO table
— still just warned, and the gate registered to close the finding claimed
the property for both.
FATAL 2, the sharpest of the three rounds: counting games proves they
STARTED. Stopping the engine after one round gives 200 games, all-zero
columns and exit 0 — byte for byte the signature CB-EV-0030 says the
instrumentation distinguishes from a real result. Both harnesses now
require every counted game to have reached an outcome over five rounds.
FATAL 3: round 2's `.csv` filter was applied to all three loops, so
catalog.yaml and rules_delta.yaml — whose missing digests were round 1's
finding — were recorded and then never compared, and never checked against
upstream at all. Only the parser loop filters now.
FATAL 4: five of six tiebreak comparators had no coverage. GR-E04's
tiebreak never executes in any scenario. All four are now covered and
mutation-verified; the Blame key needed compensating claims to be
reachable at all, since Blame also lowers the coalition score.
SERIOUS: "peak held" computed the same number as "peak assigned" for every
possible input — the real gap was that START_STRESS was an unchecked
constant, now read off the dealt state; cadence="none" was a pure
loophole, removed; sibling discovery swapped a hand-written list for
hand-written globs and missed metadata.json and VARIANT.md, both named in
the package's own changed_files — now walked, and it found them
immediately; and "~72,000 games" was unsourced, make panels runs 17,600.
Also separated two kinds of number that were presented alike: seats×games
is invariant, 363 and 29 vary 7.1%-11.5% across samples.
Round 4 owed. The conclusion is not that the work is nearly right — it is
that author-made corrections to measurement work should be assumed
defective until a fresh reader has attacked them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CB-REV-0002 #7 found the H1 harnesses were run by no gate. The fix created
a `panels` target and registered it — and did not add it to `make all`. So
the gate existed, passed lint, and ran nowhere: the same defect one level
up, committed in the fix for it.
The registry checked that a gate's target EXISTS. It never checked that
the gate RUNS, and six of ten registered gates are absent from `make all`.
Some are legitimately manual — cost-budget needs spend data, gate-review
is periodic. So a gate may be manual, but it must SAY so: every entry now
declares cadence = all | manual | none, loop-lint requires it, and a gate
claiming "all" that is not in `all:` fails. Both rules mutation-verified.
panels is cadence = all and now runs there, so the figures in CB-EV-0030
and CB-EV-0031 are regenerated by the gate rather than by hand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three FATAL, five SERIOUS. The substance of round 1's corrections held —
Reactive is genuinely one arm different, the five replacement controls are
non-inert, the inert metric is right, the numbers reproduce. What failed
were the CLAIMS about them, and two defects the corrections introduced.
FATAL 1: the fix for round 1's #11 did not fix it. The assertion was
`games + setup_fails == 200`, and a refused setup increments setup_fails
while skipping games — so the sum is invariant under exactly the failure
it claimed to catch. Injecting setup failures gave exit 0 over 196-game
columns. Now asserts games == GAMES, verified to exit 101.
FATAL 2: the correction to the selective-column FATAL was itself
selective. "81-1000 per cell, baseline AND H1" and "31-1000" twelve lines
apart, both taken from the baseline row; under H1 rank-75 arms are
59/0/0/0. Every cell is now printed rather than summarised, and the
corrected verdict is the opposite of the one it replaced: under rank-75,
H1 REDUCES DARVO arms to zero at 3p and above.
FATAL 3: "DARVO arms 2 per seat per game" is 1 per seat per game, exactly,
at every band.
SERIOUS: the tiebreak oracle asserted only that the winner set CHANGED, so
reversing the tiebreak left it green; the #13 defect's impact was claimed
and never measured (72,000 games: zero divergences — real in principle,
witnessed only by a constructed board); a 29-of-363 citation pointed at a
file that did not contain it (round 1's reviewer did report it, and it was
never transcribed — the record was wrong, not the number); the harnesses
were run by NO GATE, so every published figure came from a manual run of
an ungated binary, including the assertion added for #1; and edition-check's
sibling handling — added by the last correction — was self-certifying,
crashed instead of failing, and counted Markdown lines as coverage. Now
discovered on disk, and it found a real gap on its first run: Rules_Text.csv
vendored with no digest.
Also: "peak Stress held" was dead code kept quiet by `let _ = held;` — the
numbers were right by coincidence.
make panels is now a registered gate. Round 3 is owed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-0011 decided it: vendor the CSV with a checked digest, read it with
a ~50-line reader, and let the hashes move.
The declaration's constraint was measured against the WRONG BUDGET. It
said a CSV crate costs 21,613 against AM-4a's 3,798 of headroom, '5.7x
over, settled by measurement'. But setup and problem_priorities are
cfg(scenarios) and are not in the shipped runtime at all, so AM-4a never
sees them. Against AM-4b, csv costs 17,651 against 19,742 -- it FITS,
with 2,091 to spare. It is refused anyway, on proportion: 89% of the
budget's remaining capacity to read 20 rows. The revisit condition is
stated (nested quoting, embedded newlines, multiple dialects).
GR-S01 now deals Surface + hidden 1..=k as ruled, with edition values and
suits. Measured: 6/9/12 available against thresholds 5/7/9 -- the game is
winnable at every seat count, which is what the maintainer could not do.
gd0001 is INVERTED, not deleted, and now also asserts the 6/9/12 so a
deal that is reachable for the wrong reason still fails.
Blast radius was scenario expectations, exactly as the ADR predicted: no
scenario pinned a hash and no bundle is committed. Six scenarios and two
unit tests updated, each with a note. gr-e01-threshold-unreachable-2p is
RENAMED to -reachable- and rewritten as the non-provisional import check
ground-game asked for by name. gr-e03's setup was restructured, not just
renumbered: with values 2,2,2 its personal-edge test would have tied
three ways and asserted nothing.
BLOCKING: AM-7 fails at median 0.845 against its 0.9 floor. Isolated
across three runs -- 3 problems + stand-in 0.97, 3 problems + edition
0.909, 4 problems + edition 0.845. State is BOUNDED (proven: identical
after 5k and 100k events), so this is not the unbounded-growth defect
AM-7 exists to catch; it is a bigger working set streaming a long log.
Whether AM-7's floor is still right for a larger aggregate is a spec
question and lowering it requires an ADR, so it is not being tuned here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
T03: InnerLoop v1.7 plus loop-lint's own-cost check. Six passes
under-reported themselves by 30-45%, never once high, and the rule lived
only in evidence files having been re-derived three times. The READING
is load bearing, not the boundary: CB-WP-0018 T04 applied 're-run the
instrument at the moment of quoting' alone and its figure was correct.
So the operative instruction is re-run when you quote, and loop-lint
fails an evidence file naming its own workplan beside a dollar amount
without marking it provisional.
It binds forward from this pass. The check fires on seven historical
files which ARE the evidence for the rule; making them comply would edit
the record to remove the thing it proves -- the same category error as a
live fact: tag on a dated measurement, which this pass also hit.
Lifecycle, at the maintainer's instruction: ready -> active -> done,
where ready means declared and not started. loop-lint fails a workplan
that has started and still says ready, one that is active with
everything closed, and one that is done with an open task. The first
version of that check was WRONG and its own self-test caught it: it
stripped the leading status: assuming frontmatter, which silently
dropped a real task once the frontmatter said ready or active.
Both new checks then fired on this pass's own artifacts and both were
right to.
T04: CB-EV-0017. The new meta budget's first reading is a breach it
caused -- 27% against the 20% line, because this pass cost $31.18
against product passes averaging ~$21. Reported rather than exempted:
ADR-0006 D2 covers the instrument repairs but not the rule-writing, and
the honest reading is that this should have been two passes.
CB-WP-0018 settled at $36.53/95 against $28.08/82 last reported, 30%
higher. Seven for seven.
make all exits 0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
InnerLoop v1.7. The purpose is written first and the number follows from
it: most spend on the task at hand, some on control, review and
improving the process. make status prints it above the figure, because a
threshold with no stated purpose is what let this number be argued three
times.
Soft 20% over a trailing 5, and the self-test enforces that the ratio and
the window are a PAIR: META_SOFT_PCT == 100 / TRAILING_PASSES. One meta
pass among n at parity cost reads 1/n, so 80/20 is one pass in five at
normal cost -- a five-pass window. The same 20% over three would have
silently also demanded the meta pass be half-price, which makes meta work
rushed rather than rare. Moving the ratio without the window goes red.
The phase setting is declared, argued and expiring in gates.toml, and
reverts on review_by unless re-argued. Verified live at 35%. One with no
reason or no expiry is refused rather than honoured, because a threshold
anyone may move is not a threshold.
Measured: the last five passes read 7% against the new line.
InnerLoop.md crossed the 400-line limit three times while this was
written and was fixed structurally each time -- the arithmetic, the
cost-per-response basis and the two review case studies moved to
InnerLoopReference.md. The limit was not raised.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
T03: input::describe writes a sentence per legal command; data-descs
carries them in step with data-targets; the ghost already following the
pointer shows the one for whatever legal target is under it, so the
explanation lands beside the target with no overlay layer to keep
aligned. ADR-0010 D1 binds -- the page renders it, never composes it.
Both mutations INITIALLY SURVIVED because the fixture's Attack card had
exactly one target, where an off-by-one shift and a truncation are both
no-ops. CB-EV-0014's lesson one level in: a fixture too thin to express
a failure is how the failure survives. Two attack targets now, both red.
T04: chaos rate d4 -> d8, window 2 open at 12 declarations, retiring if
an override changes nothing twice running. Window 1's condition was NOT
met -- both overrides changed the outcome -- so the mechanism is kept.
The weakest part of the decision is that it is a rate change argued from
n=2, so window 2 carries a falsifier: no override at all is evidence the
rate went too far, not that the mechanism is healthy.
InnerLoop.md hit 401 lines and the loadability gate fired; the rationale
moved to InnerLoopReference.md, structurally, per the standing precedent
that limits are not raised.
CB-WP-0017 settled at $9.48/40 against $5.19/23 reported mid-flight,
83% higher. Six for six, always low -- read by re-running the instrument
at the moment of quoting, which is CB-EV-0015's correction applied for
the first time.
make all exits 0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Provenance (tier M, structural S, chaos d4=4 -> OVERRIDE drawn M):
the maintainer could drag after CB-WP-0016 but could not tell what was
pickable, held, or droppable. Underneath that, the page was WRONG about
which moves exist: 9 legal commands rendered as 5 cards each claiming
all three target kinds, from a const string in the emitter. Investigate
is legal on problems 2 and 3 but not 1; Solve on 1 but not 2 or 3. The
live page now says 'Solve onto problem 1'.
ADR-0010 restates control 5, which this work would otherwise have
outgrown in silence: every game fact the page acts on must arrive from
Rust as data; the script may read, match and render it, never compute,
infer, filter or default one. The survey's real finding is that the
permitted and forbidden designs are indistinguishable from outside, so
the vocabulary grep is demoted to a cheap first line and two behavioural
properties become the controls -- the highlighted set EQUALS the set
Rust emitted, and anything the page marks legal must resolve. Both
mutation-proven; the derive-legality mutation produces a plausible
highlight (seat-0,1,2 where only seat-1 is legal) and is caught.
Visible now: .pick resting shadow, .held on the grabbed element, .dropok
on every legal target including BOTH drawings of a seat, and a ghost
following the pointer. Nothing perceptual is verified and ADR-0010 D5
says so.
The DOM stub now models classList/querySelectorAll/createElement and
builds its node set from the real emitted page. Trap recorded: QuickJS
fixes its stack limit at Context creation relative to that frame, so a
helper returning a Context makes every later eval report
'SyntaxError: stack overflow'.
CHAOS WINDOW CLOSED, 12 declarations, 2 overrides, one each way. Both
changed the outcome, so the retirement condition is not met. Verdict:
keep, and recommend d4 -> d8 with a second window of 12 -- that is a
change to the loop's own constraints and is owed to the next declaration
as tier-M work, not made here.
CB-EV-0014 corrected: it quoted CB-WP-0015 at $15.14/136 and called it
the first settled figure quoted. Now $22.70/166. The number had been
read during CB-WP-0015 itself, so there are two defects -- the boundary,
and quoting from memory instead of re-running the instrument.
make all exits 0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Provenance (tier S, one paragraph in lieu of survey and ADR): the two
clauses mutation-check has reported inert since CB-WP-0005. AM-7's
scaling ratio was held up by a test literally named
replay_100k_events_is_linear_and_fast that computed both throughputs,
printed both, and never divided one by the other. AM-8's N=10 was held
up by a runner that does two.
Both are now red. AM-7 3/3, AM-8 2/2, M-D1-MUT 10/14, and ADR-0005's
>=10-of-14 prediction MET for the first time. Neither was closed by
amending the question away, which was the live risk: the denominator is
unchanged and the four unenforced rows are the four already
unenforceable.
AM-7 needed three estimators. Best-of-N per leg then divide (AM-6's,
correct for a floor on one number) gave 0.581-1.085 on an unchanged
binary; legs back-to-back gave medians 0.931-1.004; legs interleaved at
fold granularity give 0.987/0.991/0.989, and 0.989 under 8-way CPU
contention while absolute throughput fell 4x. The INDETERMINATE guard
demanded unanimity and failed a good measurement over one sample
0.001 under the floor; it now requires a two-thirds majority. The
control that matters: AM-6's constant-cost mutation halves throughput
and leaves this ratio at 0.999x green, so AM-7 is not a second AM-6.
AM-8 kept N=10 because the measurement said so. Perturbing the RNG only
from its fourth construction on: --runs 2 PASSES, --runs 10 fails. A
late-onset divergence is deterministic, not flaky, so it is a control
rather than a coin flip. Ten runs live on one scenario (make am8, ~2s)
rather than all 25 (47s a build). GameKernel 5b records it.
The full run also found AM-4a's own mutation stale since ADR-0008 D3
moved the target 250,000 -> 161,000 in CB-WP-0013 -- reported
HARNESS-BROKEN, no score published. The build-free half of that check
is now a --self-test assertion, so make all catches the next one.
mutation-check clauses may now carry their own verify and mutation, and
then the enforced flag is measured rather than declared; a declaration
disagreeing with its measurement is refused.
make all exits 0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-0008, tier M (survey and ADR merged).
D1 — SH-3 retired as a gate, kept as a diagnostic. Investigating it
found a third defect, deeper than the two this pass was declared on.
Re-deriving batching from the raw transcripts, independently of cb-cost:
CB-WP-0011 pass 54 with tools 0 batched 0.0%
gap -> next decl 16 with tools 6 batched 37.5%
CB-WP-0012 pass 86 with tools 0 batched 0.0%
gap -> next decl 10 with tools 1 batched 10.0%
CB-WP-0013 so far 10 with tools 0 batched 0.0%
Zero batched turns in 150 in-pass responses; 37.5% in one gap, above the
20% floor. Batching needs two calls whose inputs are known at once —
orientation work. Implementation consumes each step's result before the
next. SH-3's window is since the last commit, which during a pass is
always implementation. The metric could not read above ~0% in the window
it was gated on. A floor the window structurally excludes is not a
target.
This pass's own declaration was also wrong: it claimed batching "has got
worse" (7.8-8.6% vs 1.1-6.3%). Differently-placed windows, not different
behaviour. Withdrawn — the same class of error, in the pass written to
correct it.
Not retargeting to match the measurement: the floor was not moved to 6%,
the gate was removed on an argument about what the quantity is worth.
The number is still reported; only the verdict is gone.
D2/D3 — AM-4a counts --edges normal,no-proc-macro: 157,202, not 246,250.
The target moves down with it, 250,000 -> 161,000, so the correction
hands back essentially nothing (headroom 3,750 -> 3,798). Three controls:
the exclusion drops exactly the five expected crates, only removes and
never adds, and is not a no-op.
The DFD gate then caught the follow-on it exists for — three historical
documents carrying live fact tags for a number that had changed. Not
rewritten; untagged, with a supersession banner.
AM-4b is deliberately not corrected: its proc-macro share is unmeasured.
gate-review now reads 0 due, 0 silent, 0 drifted — GATE-REVIEW earns its
first caught entry by forcing SH-3's re-justification, and the registry
has no silent gates left.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CB-EV-0010. The pass's own verdict on the tier it ran at.
Full-weight review withdrew the capability port the declaration was made
to build. A tier-S pass has no step 2 and would have shipped it, and
stage 2 would have found it unimplementable — which is what CommitWindow
is already on record in this repo for doing.
Two corrections of the survey's own numbers, compounding:
AM-4a headroom 3,750 claimed -> 92,798 measured (25x)
cheapest windowed 480,501 claimed -> 140,079 measured (3.4x)
headline ratio 128x -> 1.5x (85x)
The prediction from CB-EV-0009 §4 held: meta budget reads 0%, published
in advance and unfalsified.
A correction that is now a pattern: CB-EV-0009 reported CB-WP-0011 at
45 responses / $4.23 / 0.094; final is 71 / $7.02 / 0.099. Still the
cheapest pass, so the conclusion stands. But that is the second
consecutive evidence file to report its own pass's cost low — a pass
cannot measure its own cost, and one quoting its own is quoting a floor.
First priced tier comparison on a single subject: 0.123 $/response at L
against 0.099 at S — 24% more, for a pass that found the two errors
above. On one data point, step 2 is cheap.
Not shipped, and said plainly: the emitted JavaScript has never been
executed. The socket loop is tested end to end with synthetic HTTP and
the page is asserted against as a parsed document, but no browser engine
has run it. INTENT stage 1 therefore stays open even though all four of
its named deliverables now exist.
SH-3 reads 0.0% for a sixth consecutive pass and remains the oldest
unargued number in the project.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CB-EV-0009. Tier S did not produce a worse outcome than tier L would
have. The roll deleted a survey that would have opened on 2D toolkit
selection; the decomposition it forced instead found that the existing
text renderer was showing 24 of 41 view fields. The structural trigger
fires on a property of the plan, not of the code, so nothing in the
tier derivation could have said the port was the wrong first question.
Recorded honestly in both directions: the pass also made an interface
change with no review, which is the cost side. One favourable fire is
not a calibration; the window stays open to 2026-09-30.
Corrects a number this pass asserted: the T01 commit message says
'42 of 43, up from 24'. Measured by splicing the old renderer back in,
it is 41 of 42 with 1 declared omitted, up from 24 -- and 16 fields
were genuinely absent, not 17, because outcome.winners was rendered in
a different format. Both original figures were counted by hand.
CHAOS gains its first caught entry. Cheapest pass per response yet
recorded (0.094 vs a previous best of 0.123).
ADR-0006 D3. Nine standing control gates now say what they check, what
they have caught with pointers, when their keep-or-kill argument is due,
and what would retire them. make gate-review reports what is overdue and
what has caught nothing; it never fails the build, for CB-RES-0005 §4's
reason.
Drift is checked in both directions and both are pinned by self-tests: a
dependency of `make all` that is neither a registered control gate nor
listed in not_control_gates is a loop-lint finding, so a new gate cannot
acquire permanence without a review date, and an entry naming a target
the Makefile lacks is a finding too.
First run: 0 due, 2 silent. The silent two are the chaos roll, whose
12-declaration window exists precisely to find out, and gate-review
itself, which is not exempt from its own rule — if it has retired,
tightened or forced the re-justification of nothing by 2026-12-31 it is
a ritual and goes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
