alex/sots-engine
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
sots-engine/docs
History
alex fa53e02b53 L1: probe the AI client seed across two processes -- it is fresh every time 
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
2026-09-08 18:22:14 -04:00
..
A2-alliance-and-modcount.md lane A2: S04, the alliance mask -- 80 player-records, 560 fields, 0 mismatches 2026-09-08 12:44:00 -04:00
B1.md b1: ComputeBudget hook + adapter; fix researchMoneyKept to be charged only with a research target; budget out-param is int[22] 2026-09-07 23:24:10 -04:00
B2.md b2 live: 3 completions compared 0 divergences; float32 confirmed on the game; RollResearchEvent RNG region added; xenotech block size bug fixed 2026-09-08 01:18:39 -04:00
B3.md b3 live results: 13/15 zero-divergence, RNG post-state 14/15; research target is state 3 (no decay); oracle gap = unposted overbudget event 2026-09-08 00:29:44 -04:00
B4.md b4 live: 36 calls 0 divergences, no colony RNG draws as predicted; fix 3 hook bugs (stale args, two-base StrategyServer, raw Ghidra offset) 2026-09-08 02:07:28 -04:00
B6-ship-construction.md docs: B6 measured -- prediction held, closed 0 regressed 0 on both reference pairs 2026-09-08 14:54:21 -04:00
E-events.md game/events: the player event log and the five research events 2026-09-08 02:45:57 -04:00
EV-events.md P11: post the no-research event into the save's turn bucket 2026-09-08 15:39:42 -04:00
G-wire-schema.md A: type the AIAgent custom-data blocks; named coverage 98.0% -> 99.9% 2026-09-08 07:33:06 -04:00
G3-civilian-growth.md S11 civilian growth; fix ComputeBudget's interest literals 2026-09-08 15:52:38 -04:00
game-data.md game/data: typed catalogs (weapons, sections, turrets, ids, tech tree, strings), cross_check; oracle 229k values 0 diffs 2026-09-07 17:55:08 -04:00
game-design.md game/design: assembly/fit/gating rules + derived stats on game/data; 127/127 designs, 197/197 DOpts, oracle-identical 2026-09-07 22:10:35 -04:00
game-effects.md b2: OnTechResearched hook + 10 effects-table corrections (float32, xenotech block size, no already-researched guard, sticky mask, AI bonus values) 2026-09-08 00:01:04 -04:00
game-sim.md b4: colony + movement hooks; 22 formula corrections (growth curve has no capacity term, range margin +0.05f, ties-to-even rounding); 3 verified signatures 2026-09-08 00:57:55 -04:00
H-probes.md lane H: entry-probe module, ProcessTeamRecord hook, probes= config, and the outcome of the five predictions 2026-09-08 13:38:59 -04:00
harness-audit.md lane P: post the research events from ours (count-only), declare ObservedTech 2026-09-08 04:16:22 -04:00
L1-predictions.md L1: probe the AI client seed across two processes -- it is fresh every time 2026-09-08 18:22:14 -04:00
M-movefleet.md movefleet: reproduce the original's float32 normalise; 8 of 45 live divergences -> 0 2026-09-08 03:50:04 -04:00
M0.md m0: binkw32 proxy shim, minhook, trace hook on Application::Initialize; build/deploy tooling 2026-09-07 17:22:15 -04:00
M1.md m1: GlobalConsts::LoadFile hook + game/config loader; trace 19 calls, compare 0 div, replace passes End-Turn oracle 2026-09-07 22:16:55 -04:00
M2.md m2: live verification results (weapons compare 0 div, oracle passes; section compare crash documented) 2026-09-07 23:52:50 -04:00
mars-parse.md mars/parse + mars/text: engine parity (Mars::Script tokenizer, first-wins keys, trailing-pair drop); oracles 1531/1531, 64/64 2026-09-07 18:10:18 -04:00
mars-rng.md mars::rng: the seven draw entry points, with their word costs 2026-09-08 09:59:05 -04:00
mars-stream.md mars/stream + mars/rng: Streamable reader/writer/shapes, MT19937; 3 real saves exact + byte-identical round-trip 2026-09-07 17:46:06 -04:00
mars-text.md mars/parse + mars/text: engine parity (Mars::Script tokenizer, first-wins keys, trailing-pair drop); oracles 1531/1531, 64/64 2026-09-07 18:10:18 -04:00
mars-vfs.md mars/vfs: gob ZIP reader + native override, miniz 3.1.2; oracle-verified 8352+2035 entries 2026-09-07 17:40:51 -04:00
N-output-term.md lane N: the population -> base-output term, live-verified 2026-09-08 12:11:36 -04:00
P-events-wiring.md lane P: post the research events from ours (count-only), declare ObservedTech 2026-09-08 04:16:22 -04:00
PL-players-residual.md PL: decompose the /Sim/players residual; S00 PvSav snapshot; T31 recovers its difficulty column from the save; the bankruptcy protection factor is a widened float 2026-09-08 16:34:40 -04:00
R-recapture.md recapture: golden traces with guards on the live game 2026-09-08 03:04:34 -04:00
S-standalone.md lane A2: S04, the alliance mask -- 80 player-records, 560 fields, 0 mismatches 2026-09-08 12:44:00 -04:00
shim-trace.md shim: force and verify the x87 control word around the turn gate 2026-09-08 05:09:16 -04:00
SV-script-objects.md SV: the script-object event bus, and what it writes in a turn 2026-09-08 16:46:25 -04:00
T-turn-driver.md lane T merge fixups: name the roll-succeeded branch instead of its FUN_ id (clean-room); mark describe_i32 maybe_unused so the shim cross-builds 2026-09-08 08:13:39 -04:00
U-unlock.md U: live verification - 0 divergences on 35 calls, three workloads 2026-09-08 06:57:15 -04:00
V-eventlive.md V: live verification of lane P's event posting; the advance prediction held 2026-09-08 06:00:01 -04:00
W2-predictions.md merge lane W2: watchpoints + multiplayer Tier 0 (CMake list and main.cpp union-resolved to keep both instruments; header regenerated) 2026-09-08 15:28:28 -04:00
W2-watchpoints.md merge lane W2: watchpoints + multiplayer Tier 0 (CMake list and main.cpp union-resolved to keep both instruments; header regenerated) 2026-09-08 15:28:28 -04:00
W3-predictions.md docs: name 0x00743ec0 in the W3 predictions -- clean_room_check rejected the raw identifier 2026-09-08 16:39:23 -04:00
W3-watchpoints.md W3: watch.mode=tshn -- arm the NVO/TShn record and the trade+spy containers from the same S 2026-09-08 16:38:55 -04:00
Y-standalone-rng-and-tail.md Y: correct the divisor-defect numbers -- 6 of 25 corpus records, and the rate rises with magnitude 2026-09-08 11:39:36 -04:00
Z-tail-rng.md Z: P9 outcome -- turn 64, one word, every clause held 2026-09-08 10:24:17 -04:00