Written before the game was started. Records one correction to lane L3 up front:
the turn-15 save already has a trade sector with tsnumflt=1 (the AI's), so 'any
freighter anywhere' is already falsified as the gate for the trade callees.
New shim module `ai_rng`: a bracket on StrategyClient::OnResumePlaying over the per-client
generator at +0x134, with two independent measurements per bracket (an observer on the existing
seven draw-site detours, and `left` read straight off the object) plus AI-call-site attribution
through the cl_* facades and an optional seed pin.
Measured on VM140/VM145, six runs, five fresh processes, unpinned autosaves byte-identical to the
published oracle:
turn2->turn3 human 0, AI 32 -> 3 words, AI 496 -> 0, AI 512 -> 0
turn1->turn2 human 0, AI 32 -> 7 words, AI 496 -> 0, AI 512 -> 1 word ... or 3
The last row is the answer: 512 makes ONE cl_RandRange call (the research-target tie-break at
0x006a8495, phase 18) and that one call cost 1 word in one process and 3 in another, because
RNG_NextInt is an unbounded rejection loop. The count is not fixed across clients, across turns,
or across processes with the path held fixed.
Also found: cl_RandFloat 0x00579c70, a third cl_* RNG facade that reaches RNG_NextFloat by a TAIL
JUMP and so leaves no rel32 edge for a call-graph sweep -- which is why ai-turn-logic.md 5's "zero
NextFloat calls from the AI module" reads as true when it is not.
The instrument caught its own defect: RNG_Chance calls RNG_NextFloat and both are detoured, so a
drawn word was reported twice. left_delta was never affected, which is the point of having two
measurements.
Gates run separately on a fresh build directory: clean_room_check OK, host ctest 55/55, CT111 shim
cross-build OK.
Lane RB consumed the capture and found the record was wrong: the element is dumped in
memory order and its memory field order is not its wire order. Memory member 1 is wire
member SRsc; six members are unread. RB's first replay regressed two leaves on exactly
that.
The values now ship as rates_memory_order_{u32,f32} with wire: null, and the .tcb emits
the seven '?' tokens RB's own converter emits, so a replayer refuses rather than guesses.
A missing field is honest; a mislabelled one propagates -- and a capture is meant to
outlive whatever adapter happened to read it.
Also records the outcomes against every prediction, including the one the consumer
falsified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
Lane L3 needed to watch the trade-route and spy-program vectors across a whole game
played forward, not one turn of one save. `modcount` and `tshn` arm once on purpose --
their targets are picked from one turn's state and re-picking them would move the
measurement -- so this is a third mode rather than a change to either.
`cont` puts all four debug slots on the two containers, `_Myfirst` as well as `_Mylast`.
Both vectors are default-constructed with all three pointers zero, so the first element
writes all three: watching only `_Mylast` cannot separate "allocated for the first time"
from "appended to an existing buffer", and those are different events in the model this
lane set out to falsify. It re-arms and re-logs on every End Turn, and the canary
self-test's counter is therefore read as a delta -- on the arm-once modes the delta is
the old value, so their log lines are byte-identical.
`ReportContainer` is factored out of `ArmTshnSlots` so both modes emit the same container
line. Lane W3's published count=0 is the control every later count is compared against,
and a reformatted line would have made that comparison a judgement call.
The defect: `Shim_Init` returned before `install_watchpoints` whenever the trace mode was
`off`, so `hooks=off watch=on` printed `watch=on` in the banner and armed absolutely
nothing -- a config that reports a confident zero, which is the failure method rule 1
exists to catch. The watchpoints are an independent instrument with their own arming
detour and no trace records, and a long play session wants them without paying 30-45 s
per End Turn for template hooks that measure nothing it is asking about. MinHook is now
initialised and the module installed on the `hooks=off` path when `watch=on`.
Configs: shim.cfg.l3cont / .l3control differ in exactly one key for rule 19;
.l3probe is lane H's hp11 verbatim plus the three watch keys, so the entry counts stay
comparable to lane H's empty-container baseline line for line.
Gates: clean_room_check OK; host ctest 54/54; CT111 shim cross-build OK, exports 66
names identical.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
Lane L1 found each AI client's generator takes a fresh per-process word, so the AI is
MT19937 from one word per client and game/ai is a function of (save, seed). A capture
without the seeds records the answer without the input.
- aiseed=log|pin on StrategyApp::RunAI: one detour that reads the fourth stack argument
where the callee reads it, and in pin mode replaces it. Both modes log observed AND
used, so a pinned run says so in its own capture.
- aiorders.words= widens the element window past list 1's 48-byte horizon.
- aiorders.deep= follows the heap: the route vector, the counted vector and the
Population body were ABSENT from every capture so far, not empty.
- tools/turncommands_capture.py (in sots-re) does the typing offline, so a corrected
element record costs a re-parse rather than a VM run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
The headline is one number: our turn writes ModCount 14 where the oracle writes 24, and the
missing ten are the turn's command stream. P1 decomposes the ten and names what each near
miss would mean; P5 predicts the secondary pair closes five leaves and must NOT close the
sixth, because the recording is from a run that diverged there.
The capture lane needs the block dump L4 built; branching off main without it
would mean writing the same detour twice. Header regenerated from sots-re
(rule 14), not hand-resolved: 1,217 entries.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
Two hooks, Mars::RNG::Seed 0x0049fdf0 and StrategyApp::RunAI 0x008706f0, and a
config that turns everything else off. Two launches from the same save, load
only -- the AI clients are constructed on load, so no End Turn is needed.
Result: every AI client's generator seed differs between processes (net 32,
496 and 512 all move), while the record structure is byte-for-byte the same
shape and one Seed call with seed=0 produces an identical state in both runs.
So the turn1-state -> turn2 nondeterminism is a SEED effect, not the ordering
effect that was predicted, and lane AI1's 'every draw from the static generator
returns 0' is falsified by measurement.
The prediction said the opposite and is left in docs/L1-predictions.md with its
outcome underneath.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
New shim module src/shim/hooks/ai_orders.{h,cpp}: one register-transparent entry
stub on StrategySim::ApplyTurnCommandBatch dumps every submitted TurnCommands block
(six gates, 27 list lengths, element bytes) at the point where all of them are
complete in memory; sixteen entry probes, with RunTaskList's stub recording the
pass so every later hit is pass-attributed.
Two workloads on VM145, one End Turn each. The rule-19 control passed with all
seventeen detours installed: both autosaves byte-identical to the published oracle.
What the AI actually emits, and three things no reading had produced:
- a list-23 element on EVERY turn, the first element ever observed in the free
half of the cost table -- and both turns still cost the measured 12;
- the ids in AI commands are client-allocated and travel in the command (design
18, fleet 34; neither exists in the input save);
- pass 0 emits nothing, measured from element counts rather than inferred.
tests/game_ai/test_live_blocks.cpp rebuilds both captured blocks through the public
OrderClient API and asserts the list profile, element values, gate counts and
ModCount total: 44 checks. Kept separate from test_orders.cpp, which stays the
record of what static reading predicted.
Gates: clean_room_check OK, host ctest 55/55, CT111 shim cross-build exit 0.
`ComputeBudget`'s savings-interest term is now compared against the running game at
a treasury the corpus actually contains. Three runs on VM146 from turn1-state.sav:
A (widened floats, as shipped) 3,895 calls, 0 diverged, 0 undeclared writes
B (exact decimals, the control) 2,718 calls, 1,359 diverged
The game fills savingsInterest with 499 at a treasury of 50,000, and with 380 at
38,100 -- the exact decimals pay 500 and 381. Every divergence in B lands on a
treasury that is a multiple of 100 and no other state diverges at all, which is
exactly the arithmetic. G3's rule-23 reading is now measured, not inferred, and the
one-money error is shown to propagate into `available` and `researchMoney` too.
The control also settles why the earlier 4,437-call green run was green: slot 5 IS
diffed and the harness CAN see it, so that run simply presented no boundary state.
Coverage is therefore reported as distinct states, not calls: 5 distinct treasuries,
2 of them on the boundary.
Two further rule-23 constants found in the same routine by an operand-width sweep,
corrected, and honestly marked UNVERIFIED because no reference turn can see them:
- the research-yield factor is a widened 0.85f while its two neighbours in the
same product are exact doubles. Boundary: research money a multiple of 40,000;
the run presented 9 distinct values and none is.
- the three research modifiers are summed in single precision, not double.
Boundary: two of the three non-zero; the corpus has shrm = TRM = 0.
Both are pinned by boundary cases in test_economy.cpp that fail with the decimals.
Also verified live, in the same run:
- T31's difficulty-column recovery. The live ServerPlayer+0xf9 / NPC flags on all
eight players are exactly what lane PL's save-only inversion claims, including
the awkward system-owning player that is still ambiguous because it is an NPC.
- BANKRUPTCY_PROTECTION_LIMIT_FACTOR reads 3.29999995 = (float)3.3. Its file image
is zero because the loader fills it at run time, so lane PL-3 had to assume the
value; it is now measured and the assumption was right.
Falsified, and recorded as such: the difficulty-mods record does NOT sit inline at
ServerPlayer+0x36c -- that field is a heap pointer on all eight players. The row IS
reachable from a ServerPlayer (which corrects the hook's standing coverage note),
but the fitted {3.0,1.5}/{1.0,1.0} pair remains unverified. The hook logs the
pointer and does not follow it.
The `verified` column stays 0, deliberately. Every phase this compare touches is
Partial for reasons upstream of it, and promoting one because part of it was checked
is the drift app_test_catalog exists to catch. What moved is models; see
docs/L5-live-verification.md for each one with its coverage.
Gates run separately: clean-room OK, host ctest 54/54, CT111 shim cross-build exit 0.
Lane SV recovered the script-object subsystem statically and predicted that
SVSOSwarmQueen::RegisterHives takes one RNG_NextInt per new hive inside
StrategyServer::BeginProcessTurn -- which runs inside lane Z's autosave bracket
and outside every one of its subtotals, so a draw there had never been
attributed by anything.
Five nested trace hooks, each declaring the strategic generator as a region and
each carrying a model evaluated at entry so the record can disagree with it:
StrategyServer::BeginProcessTurn the unhooked interval, plus a region
over Frame so the increment is a fact
SVSOSwarmQueen::OnTurnBegin evt 0x13, vtable slot +0x60
SVSOSwarmQueen::RegisterHives predict_new_hives from the original's
own two predicates; reads the LO/HI
config pointers live
SVSOSwarmQueen::TickHives the NextQ slip and its three gates
SVSOSlaversRefuel::UpdateDifficultyTier tail phase 20; a 4-byte region over
CDiff so a store and its ABSENCE are
distinguishable (method rule 20)
Also watch.mode=snlv: one arming line moves slot 1 from the NVO map's _Mysize to
the target system's SnLv, and the arming sweep now prints SnLv with its decoded
per-player 2-bit levels for every system, which costs no debug register.
Measured on VM140: hive creation costs 2 words in BeginProcessTurn and the
residual outside the two turn drivers is 2, not 0; the next turn it is 0 again.
LO=20 HI=30. Spica's SnLv reads 0x200 with AFlags 0. The oracle reproduced
byte for byte with every one of these detours live.
Predictions and outcomes: docs/L1-predictions.md.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
`SvSctOb` is not written by direct calls. Every update goes through an event bus: a
driver notifies the root object with an integer id, the root fans the delivery out to
every child, and each delivery is two steps -- a generic handler that takes the id, then
one event-specific vtable slot that does not. The id -> slot map is a 33-entry jump table
in the image, so which class reacts to which event is recovered and exhaustive rather than
inferred from what the saves happen to show. Five of the 33 rows are not in slot order,
including two the tail sends.
Three handlers write the eight leaves that diverged:
* the slavers' difficulty tier, on the tail's end-of-turn delivery -- a three-record
stack table scanned against the frame, boundaries 1/50/100, stored only on a change,
and at frame 100 and above the scan runs off the end and stores nothing, so the tier
can never reach 2;
* the refugees' one-shot latch, on the turn-begin delivery, with a design instantiation
behind the same latch that nothing here can do;
* the swarm queen's hives, also at turn begin, registered on the systems carrying the
SWARM's scenario tag (the queen's constructor stores 3 for that and 10 for its own
encounter id) and then ticked -- and the tick is the whole explanation of a target
turn that reads 31 after one turn and 32 after the next. It is not re-rolled; it slips
forward by one every turn the spawn gates stay shut.
New host phase H03 for the turn-begin delivery, run right after the frame counter where
the original sends it, and tail phase T20 implemented. Rules are pure in game/sim.
Measured on CT111, closed and regressed stated separately:
default turn1->turn2 209 -> 126 (was 128) closed 83, regressed 0
turn2->turn3 108 -> 67 (was 69) closed 41, regressed 0
--commit-blocked=H03 turn1->turn2 209 -> 124 closed 87, regressed 2
turn2->turn3 108 -> 67 closed 41, regressed 0
Registering a hive closes the four leaves that say which systems have hives and that they
have no queens, and opens two carrying a target turn known to be wrong: the original draws
it from the strategic generator inside the turn-begin step, outside both turn drivers, and
neither the two data-file constants nor the generator's position there is settled. That
trade is a flag, not a default.
The prediction in docs/SV-script-objects.md was committed before the build, and P5 was
wrong: it called the second pair a null control, and the second pair is where the slip
rule is tested EXACTLY -- two hives, two target turns, both landing on the oracle with no
draw and no fitting.
Gates run separately: clean-room OK, host ctest 51/51, CT111 shim cross-build clean.
The prediction text is unchanged; only the decompiler-style identifier is replaced with the name
lane E3 had already recorded for it. Rule 13's gate did exactly its job on a doc that had been
committed before the build.
Second mode in lane W2's watchpoint module. No second hook: only different arithmetic on the S the
ApplyAllTurnCommands detour already holds.
- picks the target system by predicate at arm time (AFlags == 0, NVO non-empty) and logs all 28
systems, so the choice is auditable rather than a hard-coded pointer;
- probes both candidate ServerSystem bases and logs how many systems validate under each, which
settled a documentation dispute (+0x274: 9, +0x26c: 0) by measurement;
- prints the trade-route and spy-program vector triples, which is the workload-confirmation
instrument two earlier lanes lacked.
Rule 19 control passed: the armed run reproduced the determinism oracle byte for byte.
Closed 2 on the reference pair and 4 on pair 2, regressed 0, with no operator
input. Measured, both pairs, never netted.
- S00 stamps PvSav from Sav on every live player, before any phase can move it.
0 closed on pair 1 (a no-op there), 2 on pair 2.
- T31 recovers the per-player difficulty column by recomputing BnkEl from the
colony state the input save was written from and comparing against the BnkEl
the save carries. That removes the --ai-player flag as a blocker and turns the
phase's self-check into a real one: it used to compare its POST-turn result
against the PRE-turn stored value, so its 6-of-8 only ever covered the six
players whose limit does not move. The load-time check passes for every live
player of all eleven corpus saves. T31 Blocked -> Partial; BnkPr still needs
the tuning constant.
- BANKRUPTCY_PROTECTION_LIMIT_FACTOR is multiplied in as fmul dword ptr, so it
is a float32 in the image; the engine narrows it now. Zero leaves move on this
corpus -- all seven of its BnkPr records land where the two constants agree --
and three hand-written test expectations moved (rule 23).
docs/PL-players-residual.md carries the decomposition, the predictions written
before the build, where they were wrong, and the ranked remainder.
Reference pair turn1->turn2: 81 leaves closed, 0 regressed (was 78/0).
Pair turn2->turn3: 39 closed, 0 regressed (was 36/0). With
--commit-blocked=T31 --ai-player 1: 83/0 and 41/0.
game/sim/colony: GrowCivilianPopulations models ServerSystem's civilian
growth sub-pass. The whole system's delta is clamped to 20,000,000 -- an
int64 column of the population-type table, built in the executable from
its own literals -- and on both reference pairs that clamp, not the growth
curve and not any carrying capacity, is what decides the value: the
uncapped delta is 7.5x it and the capacity headroom 25x it. So the pass
commits with no tuning table loaded, and says by how much each unmodelled
input would have to be wrong before it mattered.
The one input genuinely off the wire is the per-species civilian capacity
factor. It is handled by running the pass twice, once with the modelled
capacity and once with the system's own wire-known dcs limit, and
committing only when the two agree. Imperial growth is deliberately NOT
committed: it is a no-op on this corpus and would need a capacity the
corpus can bound from below but not from above.
game/sim/economy: both interest rates in ComputeBudget are WIDENED FLOAT
literals, (double)0.01f and (double)0.15f, and are then truncated -- so a
treasury of exactly 50,000 earns 499, not 500. This module used the exact
decimals, which left the human's savings one money high on the first
reference pair and exact on the second. Sixteen hand-computed test
expectations moved by one; they were derived from the model, not measured.
The live ComputeBudget compare (4,437 calls, 0 divergences) did not catch
this because it presented only 20 distinct states and none sat on a
rounding boundary.
game/sim/colony: ShipRepairCost, the last unmodelled input of the output
turn path. The demand is still 0 -- its two design fields are cached stats
the save does not carry -- but the zero is now evidenced rather than
silent: S13 reports the candidate set, and the independent colony keeps a
ten-ship fleet over a colony whose savings close exactly at zero demand.
Gates run as separate commands: clean-room OK, host ctest 49/49, and the
CT111 shim cross-build exit 0 (required: game/sim is compiled into the
shim). The host build and report were also re-run on CT111 and produced
identical numbers.
docs/G3-civilian-growth.md; notes repo
findings/subsystems/population-growth.md.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
The engine has modelled the event log since lane E, but the standalone never wrote
any of it into the save it produces. This wires the two together for the one event
on the reference pair the standalone can compute, and corrects the condition.
What a turn actually posts, measured over all eleven saves: two events on
turn1 -> turn2 and three on turn2 -> turn3, and only two players in the whole
corpus ever hold an event at all. Order is readable off the ids -- the build pass
posts before the research pass. See docs/EV-events.md section 1.
The condition had two of the original's three tests. The missing one is "no tech
finished on this turn or later", and it is what keeps the event off the turn a tech
lands; zuul-turn23 exercises it. The third input, whether the player holds a target
at the moment of the check, is not on the wire -- three of the four real players
acquire one during the turn, which is AI research selection -- so the operator's
--ai-player roster stands in for it as a stated hypothesis and the phase stays
blocked without it.
No prose in the engine: the record's text is resolved through a caller-supplied
lookup over the operator's own installed string table, and a run without a data
root posts nothing rather than writing a record it cannot fill.
Measured (CT111 host build, real data root), closed and regressed never netted:
turn1 -> turn2 209 -> 127 closed 82 (+4 this lane), regressed 0
turn2 -> turn3 108 -> 69 closed 39 (+3 this lane), regressed 0
The posted record agrees with the oracle on all eight of its fields. Controls: with
the roster withheld the phase over-fires and regresses 16 leaves on pair 1; with the
string table withheld it posts nothing and regresses none.
Gates run separately: tools/clean_room_check.sh OK; host ctest 50/50; CT111 shim
cross-build OK (exports 66 names identical to binkw32.dll, staged in
/srv/re-lab/shim/dist-ev); CT111 host ctest 50/50.
209 -> 158 (closed 51, regressed 0) and 108 -> 87 (closed 21, regressed 0), identical to the
pre-lane baseline, measured on CT111. Falsification hypotheses 1 and 2 refuted by measurement
(three empty BQ frames per reference save, every hbq false, the one command block empty);
hypothesis 3 refuted by the corpus oracle; hypothesis 4 stands as a labelled hypothesis.
construction.h also carries the ship/fleet birth shape as read but NOT implemented, with the
reason stated: the newborn hull copies four cached stat words the engine does not yet compute.
Written before any code (earned rule 2). Both reference saves carry three BQ frames with
zero orders, every hbq false and the only TurnCommands_v5 block empty, so the pass has
nothing to advance; the archived destroyer is created inside the turn by the AI. Includes
the falsification list and names the oracle pair the campaign already owns.
The spine's fourth phase, read byte-for-byte and implemented:
almem[i] = (1 << i) | (ALid != -1 ? AL : 0)
with i the player's POSITION IN THE PLAYER VECTOR, not its index field. Both
inputs are on the wire and so is the output, through the turn-record archive,
so the phase is checkable against bytes the original wrote:
app_turn_record: 11 saves, 80 player-records, 560 fields, 0 mismatches
(was 480 fields over six fields; almem is the seventh)
The eight zero masks of the corpus's earliest archived turn are PREDICTED, not
excluded: the archiving phase also runs on load, and the load path does not run
the spine. BuildTurnRecord takes spineRan and models it, so all 80 records are
compared.
Three parts of the rule the corpus cannot separate -- the bit index, the OR,
and the ALid guard -- are pinned in app_alliance with the separating inputs no
save provides, and app_turn_record prints that it could not separate them.
Divergence, closed and regressed reported separately:
turn1->turn2 default 209 -> 204 closed 5, regressed 0
turn1->turn2 --commit-blocked 209 -> 189 closed 29, regressed 9 (was 17)
turn2->turn3 default 108 -> 103 closed 5, regressed 0
turn2->turn3 --commit-blocked 108 -> 106 closed 13, regressed 11 (was 19)
T36 stays blocked: nine leaves would still be wrong (inc x3, sav x3 behind the
budget; three census leaves behind the design catalogue). It now closes all 24
turnstats leaves on the reference pair, so it becomes a clean +24 once those
two land.
Prediction and falsification committed first in 49ae628.
Gates run separately: clean-room OK; host ctest 43/43. No src/shim touched.
Reads the whole colony output chain off the instruction stream (every range
disassembled to the next function start) and compares two of its functions
against the running game.
The population -> output law is linear and is carried by the executable:
output points per head are typeOutputModifier x 1.8 / 500000, and the
three-row population-type table is built in code rather than loaded, so the
imperial (1.0) and civilian (0.33f) modifiers are facts about the binary.
A system's total output is a SUM of three terms, not one multiplicative
chain. The station bonus scales only the imperial term and morale only the
civilian one, so OutputModifiers no longer carries either; they belong to
GroupOutputInputs. The function previously described as the base-output term
is the over-harvest RESOURCE demand, and it is corrected in place.
Live on VM140, both hooks in compare mode over two species and two workloads:
GroupOutput 13,105 calls / 0 divergences; ComputeTotalOutput 11,252 calls /
1 divergence of one ulp, in a value its caller rounds to an integer. Both
functions declare a whole-object Guard: 0 undeclared writes in 24,357 calls,
which is what makes the side-effect-free claim a measurement.
sim::Narrow forces the double rounding a 32-bit x87 build otherwise skips;
without it every civilian row came out one ulp low.
Also fixes ComputeBankruptcyLimits' elimination divisor, which was the
decimal -0.15 rather than the image's widened float -0.15000000596046448.
The two disagree for every maximum income divisible by 3 and for essentially
every empire above ~3,000,000.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
