Seed lease map for the exploratory workspace

This is the authoritative allocation for exploratory work in this session. Every range here is development or training data: once read, it can never become confirmation evidence. No protected (0x7d……) or final (0xd7……) seed has been opened.

Why these prefixes

Candidate prefixes were chosen by extracting every eight-hex-digit constant from docs/research/history.md, approaches/, src/, research/ and artifacts/ and selecting prefixes that appear nowhere. Historical gameplay seeds cluster densely in 0x3d000000–0x3ea00000, with further blocks at 0x2d, 0x2e, 0x3f, 0x4d, 0x5d700000, 0x6d, 0x6e and 0xa5700000. The only 0xa5 constants anywhere in the repository are 0xa511e9b3, 0xa54ff53a and 0xa5700000, none of which fall in the ranges below.

Allocation

Lease	Range	Games	Role	Consumer	State
SEEDLEASE-A51D eval	0xa51d0000–0xa51d003f	64	development, paired evaluation	fair-D4 reference cohort, terminal-utility sweep (6 arms), chance-strata arms (3 and 7)	opened
SEEDLEASE-A51D parity	0xa51d0100–0xa51d0101	2	development, CHECK probes	parameterized-search parity gate	opened
SEEDLEASE-A51D smoke	0xa51d8000–0xa51d800f	16	development, smoke test	corpus-generator smoke run	opened
SEEDLEASE-A51D withdrawn	0xa51d9000–0xa51dcfff	16,384	withdrawn — see incident below	first training-corpus attempt	opened, output destroyed
SEEDLEASE-A51D-SCEN	0xa51dc000–0xa51dffff	16,384	development, scenario minting	scenario engine, exact solver, minted suites	opened
SEEDLEASE-A51D-VETO	0xa51e0000–0xa51e3fff	16,384	development, paired SCREEN	rollout-veto retest	opened
SEEDLEASE-A52 d2	0xa5200000–0xa520ffff	65,536	training	corpus, depth-2 behaviour, ε=0.08	opened
SEEDLEASE-A52 d1	0xa5210000–0xa5213fff	16,384	training	corpus, depth-1 behaviour, ε=0.15	opened
SEEDLEASE-A52 d3	0xa5214000–0xa5214fff	4,096	training	corpus, depth-3 behaviour, ε=0.05	opened
SEEDLEASE-A52 d4	0xa5215000–0xa52152ff	768	training	corpus, depth-4 behaviour, ε=0.03	opened
SEEDLEASE-A51D confirm	0xa51d1000–0xa51d103f	64	development, paired evaluation	shared evaluation cohort — fresh-seed 7-strata confirmation, depth x strata factorial, learned-leaf 2x2	opened
SEEDLEASE-A52-FLOW	0xa5230000–0xa5233fff	16,384	development, diagnostic	flow-ceiling clairvoyant planner	opened (8 seeds read)
SEEDLEASE-A52-FLOW2	0xa5234000–0xa5237fff	16,384	development, diagnostic	fair planner, information-vs-planning decomposition	opened
SEEDLEASE-A52-FLOW3	0xa5238000–0xa523bfff	16,384	development, diagnostic	fair-planner ceiling extension (K=1024)	opened
SEEDLEASE-A52-LEAF	0xa5240000–0xa5247fff	32,768	development, tuning	learned-leaf blend tuning (disjoint from evaluation)	opened
SEEDLEASE-A52-REVEAL	0xa5250000–0xa5257fff	32,768	development, tuning	reveal-vs-next-disc sampling split	opened
SEEDLEASE-A52-SUITE	0xa5258000–0xa525bfff	16,384	development, diagnostic	scenario-suite validation, position mode	opened
SEEDLEASE-A52-DISTILL	0xa5260000–0xa526ffff	65,536	training	fair-planner distillation corpus	opened
SEEDLEASE-A52-FAST	0xa5270000–0xa5277fff	32,768	development, equivalence gates	semantics-preserving engine optimisation; C++ fast engine used 0xa5270000–0xa5275fff, Rust bitboard engine used 0xa5276000–0xa5277fff (gates …6000–…6fff, benchmark …7000–…7fff)	opened
SEEDLEASE-A52 reserve	0xa5216000–0xa522ffff and 0xa5278000–0xa52fffff	—	unallocated	—	reserved
SL-…-5da70000	0x5da70000–0x5da70fff	4,096	public-development	a concurrent agent's afterstate track	state disputed — under review
SL-20260822T020000Z-a5290000	0xa5290000–0xa529ffff	65,536	training	leaf-evolution CMA-ES fitness blocks, 32 seeds per generation from 0xa5290000 (at most 40 generations authorised: 0xa5290000–0xa52904ff)	opened
SL-20260822T020000Z-a52b0000	0xa52b0000–0xa52b00ff	256	public-development, held-out SCREEN	leaf-evolution held-out paired screen (first 64 seeds, read once after the candidate is frozen); remainder reserved for a replication	reserved
SL-20260822T060000Z-a52c0000	0xa52c0000–0xa52c00ff	256	public-development, SCREEN	survival-instinct root-filter screen (first 128 seeds, read once after CHECK gates); remainder reserved for replication	reserved
SL-20260823T100000Z-a52d0000	0xa52d0000–0xa52d01ff	512	public-development, SCREEN	reveal-construction leaf screen — corpus gate failed, never opened	cancelled-unopened
SL-20260823T110000Z-a52d0200	0xa52d0200–0xa52d03ff	512	public-development, SCREEN	reveal-construction leaf successor screen (first 256 seeds read once: frozen, A900 full; A300, B stopped at 39/36 games); remainder reserved	opened
SL-20260823T200000Z-a52e0000	0xa52e0000–0xa52e01ff	512	training, CHECK diagnostic	H-pool stage D0: 64 oracle games (0xa52e0000–0xa52e003f) and fair-D4 matched pool games (0xa52e0100–0xa52e01ff); public states exported, privileged generator only	reserved
SL-20260823T215000Z-a5216000	0xa5216000–0xa52191ff	2,816 used of 12,800	training, CHECK	P-SOL-1 sibling-outcome corpus: G0 ladder 0xa5216000–0xa52160ff, main corpus 0xa5217000–0xa52177ff, gate set 0xa5219000–0xa52191ff (gate origins development-read on use)	reserved
SL-20260825T063000Z-a52e0300	0xa52e0300–0xa52e12ff	4,096	training	NNUE-evolution: teacher corpus 0xa52e0300–0xa52e04ff (512 d5s7 games), evolution fitness 0xa52e0500–0xa52e0c7f (60 generations × 32 paired games), elite re-selection 0xa52e0c80–0xa52e0cff (128 games); remainder reserved	reserved
SL-20260825T063000Z-a52e1300	0xa52e1300–0xa52e14ff	512	public-development, held-out SCREEN	NNUE-evolution held-out paired screen (first 64 seeds, read once after the candidate is frozen); remainder reserved for a replication	reserved
SL-20260905T191317Z-09895ea2	0xa5300000–0xa54effff	2,031,616	training	ntuple-scale TD training games (pilot arms and main run), read in order and wrapped	opened
SL-20260905T191317Z-549265de	0xa52f2240–0xa52f227f	64	training, validation	ntuple-scale paired validation line-up at every validation point; chose the configuration and the candidate	opened
SL-20260905T191317Z-c25f58cd	0xa52f2140–0xa52f223f	256	public-development, held-out SCREEN	ntuple-scale one-shot screen, four arms on identical seeds	opened
SL-20260906T013222Z-ce319097	0xa5500000–0xa56effff	2,031,616	training	ntuple-scale replication at larger scale: TD training games of the wide layout, read in order and wrapped; stops below the historical 0xa5700000 block	opened
SL-20260906T013222Z-6cce6192	0xa52f2280–0xa52f237f	256	training, validation	ntuple-scale replication: paired validation line-up at every validation point; drives the plateau stop rule and the candidate choice	opened
SL-20260906T013222Z-f34b23c0	0xa52f2380–0xa52f257f	512	public-development, held-out SCREEN	ntuple-scale replication one-shot screen, six arms on identical seeds including the first experiment's frozen tables	opened
SL-20260906T081306Z-1aaec419	0xa52f2580–0xa52f277f	512	public-development, held-out SCREEN	ntuple-scale depth-4 screen of the frozen tables, four arms on identical seeds; no training	opened
SL-20260906T201104Z-88b984ce	0xa5800000–0xa59effff	2,031,616	training	fill-conditioned leaf: TD training games of the three warm-started arms (control, occ5, hgt5), read in order and wrapped; above the historical 0xa5700000–0xa571869f block	opened
SL-20260906T201104Z-26371f8b	0xa52f2780–0xa52f287f	256	training, validation	fill-conditioned leaf: paired validation line-up at every validation point of every arm; drives each arm's plateau rule, its best point and the candidate arm	opened
SL-20260906T201104Z-53350936	0xa52f2880–0xa52f2a7f	512	public-development, held-out SCREEN	fill-conditioned leaf one-shot screen, eleven arms on identical seeds (frozen tables, fill candidate, control, two edits, at depth 3 and depth 4)	opened

Training and evaluation ranges are disjoint by construction: evaluation lives under 0xa51d, training under 0xa52. No model trained on SEEDLEASE-A52 has ever seen an evaluation seed.

Incident: lease overlap, 2026-08-20

What happened. The first training-corpus run was launched at 0xa51d9000 for 16,384 games. 0xa51d9000 + 16384 = 0xa51dd000, so it consumed 0xa51d9000–0xa51dcfff and overlapped the scenario-minting sub-lease 0xa51dc000–0xa51dffff by 4,096 seeds (0xa51dc000–0xa51dcfff). The range was sized in games and the end address was not checked against the allocation table before launch.

Why it mattered. Both roles are development data, so no protected or final seed was touched and no qualification claim is affected. The real hazard was contamination of a benchmark: scenario positions minted from those 4,096 seeds could have appeared in the training corpus, so a model trained on that corpus would have been evaluated partly on positions harvested from its own training games. That is precisely the leak design-01-benchmark-suite.md exists to prevent.

Resolution. The contaminated corpus (d2eps08.states, 1,029,206 records) was destroyed, not filtered, and the corpus was regenerated from SEEDLEASE-A52, a range with no overlap with any evaluation or scenario lease. Regeneration cost 103 seconds. The 4,096 overlapped seeds remain validly allocated to scenario minting; they are recorded above as withdrawn from training so that no future run reuses them for that purpose.

Process change. Every block in SEEDLEASE-A52 starts exactly where the previous one ends, the arithmetic is printed and checked before launch, and the generation script records the consumed end address for each block.

Incident 2: sub-leases declared outside this file, 2026-08-21

What happened. Nine sub-leases were issued to delegated agents in their task briefs and recorded only in those agents' own READMEs and finding headers. They were absent from this table and from research/seeds/leases/, so for several hours the authoritative allocation map was incomplete. An independent reconciliation pass found four of them; the remainder were added at the same time.

Why it mattered. No collision resulted — the ranges were issued sequentially from a single coordinator and are disjoint, and a check against the concurrent OpenCode contributor's ranges (0x5da70000–0x5da70fff and 0x5eed0001–0x5eed0008) confirms no overlap in either direction. But the protection was luck plus sequential issuance, not a record. The first incident in this file was caused by exactly that failure mode: arithmetic done at launch time rather than checked against a table.

Resolution. Every range is now listed above. Machine-readable research/seeds/leases/ records remain outstanding for the 0xa5…… family and are owed before any of this work is promoted beyond the exploratory namespace.

Process change. A sub-lease is not issued until it appears in this table.