{
 "notes": "EP10 (10 of 10, the last): FINISH. Handed 30: confirm-k01 (8) and the 22 t6 legs. No new batches (no episode could read them); t5-a, t5-b, t5-e cancelled (their waiting items would spend seats nobody reads; the 19 t5 results already in goal.json are in `read` for the write-up). Items: 193 of 200 used, 7 unused.\n\nCONFIRMED FILE (the requester's rule: > 2 se over IB on fresh blocks, >= +0.3 on >= 2,048 seeds): k01 = https://api.imdnotdead.fun/artifacts/12606d82ec57ee9fe351945ec520b1ab8d79c8a472ff46b3bce6c5c0693b7b49 (code sha256 9253b6f3a49eafc1ed21ad19d9e2c0761c2ff151a6a7aa81e16de83834ad6a1b, port 732c0fb6..., `params` = W1, b1, W2, b2, 41 constants by name, scaling; `code_zb64` its Solidity; T03's structure with only literals changed). confirm-k01 k01-c1..c6, six fresh 512-seed blocks 8174080..8177151, pooled by sum/sumsq: k01 vs IB n 3,072, mean_diff +0.822, se 0.016 (50 se); per block +0.757, +0.826, +0.816, +0.847, +0.841, +0.847; k01 vs T03 (T03 arm on the same seeds) +0.068 se 0.009 (7.9 se); T03 vs IB +0.754 se 0.016. Quartiles of IB's retail volume (q1 quietest, 768 each): +0.764, +0.794, +0.816, +0.914. worst5 vs IB: 8175538 -3.06, 8175543 -2.83, 8174161 -2.53, 8174340 -2.49, 8174709 -2.48; best5: 8176213 +11.46, 8174145 +10.69, 8175552 +6.74, 8176731 +6.35, 8176476 +6.28. Gas: k01 afterSwap peak 235,550..238,271 across items (IB 234,872..237,699), 0 OOG, 0 reverts in about 5.5 M afterSwap calls per arm per item; valid in all six. Volume deltas vs IB (c1): retail +340 Y, arb -67 Y per market; the engine exposes no retail/arb edge split. k07 (t3-b; id 9212fa7f..., code 039fa1a3...): k07-c1/c2, 8177152..8178175, n 1,024: +0.884 se 0.049 over IB (c2 has one large seed, se 0.091), +0.073 se 0.015 over T03. Both are files the requester may submit; k01 has the larger confirmation.\n\nBEST TRAINED FILE (round 5, confirmed on 512 real seeds only): q22 = t6-e/q22-k06-s03-h60-p16, https://api.imdnotdead.fun/artifacts/8cbbcd23417da9dc12595527212316c61f922c56d60bbfbdfe7b65d61488c91a (fetched, hash equal; 39,853 bytes): code_sha256 97e538c7302a6d5af28db53408595358ce4fb1085714232110c295c032e2c993, port_sha256 4c2c3549cba6539a3a1b89baf76e107d0da57a713c8c8e3eeabcf35b50503f36, `params` (W1, b1, W2, b2, constants, scaling), `code_zb64`, `port_zb64`; START k06 (e7b20742..., code ef01fb90...). 32 CPUs, 3,557 markets/s, 1,371 generations in 66.6 min (pop 16, sigma0 0.3 cap half-life 60, 0.05 rms / 2^(0.03 z), M 512, train 268000000..268701951, ES seed 20261622), monitor +0.061 at gen 1,358 still rising. Gates: A 6.2e-5 bps on recorded callbacks 9200000..9200049, B +0.846 vs real +0.792 (error 0.054), C passed. Holdout 41090000..41098191 (once): +0.050 se 0.005 over k06 (retail -0.61, arb +0.66), +0.895 se 0.010 over IB (retail -0.36, arb +1.26); k06 +0.845 over IB on the same seeds. Real 8199680..8200191 (n 512): +0.911 se 0.043 over IB (quartiles +0.80/+0.94/+0.91/+0.99; worst5 8199971 -2.02, 8200084 -1.25, 8200060 -0.86, 8199933 -0.76, 8200102 -0.74), +0.059 se 0.023 over k06 (2.6 se), k06 +0.852 over IB; mean edges IB 529.45, q22 530.36. afterSwap peak 237,613 (IB 237,052), 0 OOG, 0 reverts; valid. Recommend the requester confirm q22 on 2,048 fresh seeds from 8200704 (never used) before submitting it; until then k01 is the confirmed deliverable and q22 the best candidate (about +0.05..0.09 over k01: holdout 0.895 vs 0.848 on the same seeds, not paired against k01 on the real engine).\n\nOTHER ROUND-5 LEGS (holdout 41090000..+8191 shared; frozen_vs_start se about 0.005; real n 512 se about 0.04; all valid, 0 OOG, peaks 235k..238k):\n- q04 (t6-a, k01 child, macOS replica_only, 1,887 markets/s, 1,114 gens): +0.045 se 0.005 over k01, +0.893 over IB on the holdout; NO Solidity, no gas, no real; params in its result (id 5922e652490779e5b5da5d0477b18e36577e101006f8ce19678d626a4e517c4f, port 732c0fb6). Unfrozen; a Linux freeze would be needed.\n- q14 (t6-c, k01 child; id c05bc26a8ed836f240003487d23044d252cd6ab9e7a9a782743f5347fef235cd, code 7bc0171f...; 573 gens, pop 24): holdout +0.042, +0.890 over IB; real 8191488..+511 +0.866 se 0.037 over IB, +0.028 se 0.022 over k01; peak 237,479.\n- q21 (t6-e, k06 child; id 711657aa..., code c4805881...; 397 gens): holdout +0.032, +0.878; real 8198656..+511 +0.889 se 0.039 over IB (START arm not run).\n- q13 (t6-c, k01 child; id 89d4f522..., code 08ec8003...; 300 gens): holdout +0.026, +0.874; real 8190464..+511 +0.899 se 0.034 over IB, +0.037 se 0.019 over k01.\n- q01 (t6-a; 322 gens): holdout +0.021, +0.869; real +0.800 over IB, -0.006 se 0.016 over k01.\n- q10: -0.002 holdout; the other 15 legs returned START unchanged (nothing beat START on the monitor): 68..300 generations on seats at 0..550 markets/s. Their real arms add START-vs-IB blocks: k01 +0.74..+0.90 (n 64..768), k07 +0.77..+0.90, k06 +0.77..+0.82, all consistent with the confirmations.\nPattern of the whole relay: gain over START tracks generations (1,371 gens +0.050; 1,114 +0.045; 573 +0.042; 397 +0.032; 300 +0.026; under 280 nothing), i.e. the seat's speed, with 45..60-minute sigma half-lives best and 90 worst. q21's first phase lost -0.13 by minute 20 because 2^(0.03 z) on decay constants near 1 (TOX_DECAY, LAMBDA_DECAY) is destructive; legs that capped decays at 0.999..0.9995 gained, and that cap binds (q04, q22 sit at 0.999).\n\nCURVE BY ROUND (real paired vs IB): IB 0; J5 +0.55 (n 5,120, requester); J11 +0.60 (512); T03 +0.797 (3,072: own 1,024 + confirm-t03 2,048; +0.754 on confirm-k01's 3,072); k01 +0.822 (3,072); k07 +0.884 (1,024); q22 +0.911 (512). Head-only continuations of IB, FA, IG gained +0.01..0.06; joint searches (178 weights + 41 of T's constants) carried every step from J5 on. Gains since T03 are mostly arbitrage (replica split), retail flat or slightly negative.\n\nGATE A STORY for the write-up: T03 and every descendant (k01, k06, k07, q22 share its Solidity with only literals changed) revert with Panic(0x12) only on synthetic stress paths (reserve X drained to 0.067, spot 2e8; _band's acc[0] = 0); 0 reverts in every real-engine run (tens of millions of afterSwap calls); T03G (one guard line, code 26c960d8..., dz-t03/dz-b) is identical per seed on 512 real seeds; per requester note 7 the code ships unchanged. Gates since EP8 replay recorded replica callbacks (seeds 9200000..9200149), < 1e-4 bps on every leg.\n\nSEAT MIX: 2 to 32 CPUs, 44 to 3,557 replica markets/s; macOS arm64 seats ran replica-only (q04) or blocked in early rounds; seats under 100 markets/s ran confirmations. About two thirds of trained legs across the goal came back with START unchanged.\n\nNOT CHECKED: no leaderboard run of any file; q22, q14, q21, q13 confirmed on 512 real seeds only (no 2,048-seed confirmation); q04 never frozen to Solidity; no retail/arb edge split on the real engine (replica split and one leg's reconstruction only); the 19 t5 results (T03 and k20 children, holdout 41070000) were never read by an episode and are listed in `read`; the sister goal's numbers (J5 on 5,120 seeds) are the requester's. Seeds: real below 8200704 used, plus 8588000..8589023 and 8700000..8702047; next free 8200704. Replica train used 100000000..275999999, 500000000..516000000, 600000000..800000000, 900000000..1360000000; monitor 30000000..30002047 reused on purpose; holdouts 41010000, 41030000, 41050000, 41070000, 41090000 (+8191 each).\n\nWRITE-UP: name k01 as the confirmed file (URL, code sha256, params location, n 3,072 +0.822 se 0.016, worst5/best5, quartiles, gas) and q22 as the best trained file pending a 2,048-seed confirmation; cite confirm-k01/k01-c1..c6, k07-c1..c2, t6-e/q22, t6-c/q14, t6-e/q21, t6-c/q13, t6-a/q04, confirm-t03, dz-t03/dz-b, t3-a/k01; state what is missing as above.",
 "read": [
  "confirm-k01/k01-c1",
  "confirm-k01/k01-c2",
  "confirm-k01/k01-c3",
  "confirm-k01/k01-c4",
  "confirm-k01/k01-c5",
  "confirm-k01/k01-c6",
  "confirm-k01/k07-c1",
  "confirm-k01/k07-c2",
  "t6-e/q22-k06-s03-h60-p16",
  "t6-c/q14-k01-s03-h60-p24",
  "t6-e/q21-k06-s04-h45-p16",
  "t6-c/q13-k01-s02-h45-p16",
  "t6-a/q04-k01-s03-h30-p16",
  "t5-c/n11-s07-h45-p16",
  "t5-c/n12-s04-h60-p16",
  "t5-c/n13-s05-h75-p16",
  "t5-c/n14-s06-h45-p16",
  "t5-c/n15-s03-h90-p16",
  "t5-d/n16-s03-h60-p16",
  "t5-d/n17-s04-h60-p16",
  "t5-d/n18-s05-h45-p16",
  "t5-d/n19-s03-h90-p16",
  "t5-d/n20-s04-h45-p24",
  "t5-e/n21-s05-h60-p16",
  "t5-e/n22-s04-h75-p16",
  "t5-e/n23-s03-h45-p16",
  "t5-e/n24-s06-h60-p16",
  "t5-a/n02-s07-h45-p16",
  "t5-a/n04-s05-h45-p24",
  "t5-a/n05-s03-h60-p16"
 ],
 "cancel": [
  "t5-a",
  "t5-b",
  "t5-e"
 ],
 "finish": true
}