From 7868be22df91b05331009236a89d09edbcd5b056 Mon Sep 17 00:00:00 2001 From: igneum-labs <337424239+igneum-labs@users.noreply.github.com> Date: Thu, 8 Oct 2026 21:09:00 +0000 Subject: [PATCH] Igneum 2.0 register: the D5 row's live line (the epoch cut as a partition, FAILED on 2.0.1, the fix software rolled); Counter ASIC 3.0 status: the first post-roll lock, the second fresh-install class, the freeze tree's pairing and the signing id 2a1d6caab4c24564, the chain-paired packs, the prover decision, GOV-03 PASS, the succession PASS with the H and W proposal, the night rule and Phase 1, the founder's 21:33 word, the four-chain split and main's ruling Co-Authored-By: Claude Fable 5.1 --- docs/plans/counter-asic-3-status.md | 28 ++++++++++++++++++++++++++++ docs/plans/igneum-2.0-register.md | 2 +- 2 files changed, 29 insertions(+), 1 deletion(-) diff --git a/docs/plans/counter-asic-3-status.md b/docs/plans/counter-asic-3-status.md index 3627be2ae..942781d36 100644 --- a/docs/plans/counter-asic-3-status.md +++ b/docs/plans/counter-asic-3-status.md @@ -565,6 +565,34 @@ THE NIGHT RULE (main's words, the founder to bed, 21:3x UK): (1) From this line A FAULT FOUND AT 21:28 (the build-server lane by pid; the coordinator's read at 21:29): a dl-host deploy running from the Mac with no recorded pid, `node /Users/joshm/Projects/igneum/tools/workers/push.mjs` (pid 47013, parent launchd, started 21:28:00; npm exec vercel 50070, cwd igneum-dlsite), re-hashing the 5.9 GB dl/ tree each fire: the launchd agent com.igneum.workers-push (~/Library/LaunchAgents/com.igneum.workers-push.plist), the old Mac workers publish still scheduled after the hand-over to build-1; the class the founder named; the fix: the agent unloaded by its label and its plist retired by whoever installed it (the site lane), never a kill by name; the build-server lane's one Arc deploy serialized behind it with its pid recorded. THE LAUNCHD AGENT UNLOADED (21:30 UK, the site lane, read back): com.igneum.workers-push unloaded by its label (launchctl bootout gui/501/com.igneum.workers-push), its plist renamed .retired-20261008, launchctl list no longer showing it; its running upload (pid 47013) had ended on its own, nothing killed; nothing publishes from the Mac now but tools/site-deploy-from-mirror.sh with its pid file. Candidate (c) e8773ff5 all six suites green on the final tree (184/143/74/30/38/19), the pair building, then the canaries and the id read-back. The night rule broadcast to every lane at 21:3x with its stay or release. +MAIN's ADDITION TO THE 08:00 READ-BACK (21:3x): the accepted-block count since the devnet-4 genesis for each of the mini, PC 1 and PC 2, the time each first mined on 2.0.1 or 2.0.2 and its current rate, from their logs through the relay and the mini's api, the first line (the founder noting they have mined almost nothing since 2.0); the launchd workers-push agent confirmed retired by main's own read. + +THE FOUNDER'S WORD at 21:33 UK: "Do not let anything die or bug out overnight. I want to wake up to serious progress. Build at maximum speed." Two amendments to the night rule (the messaging rule unchanged: real changes only, one line an hour from the coordinator, the 08:00 read-back). (1) NOTHING DIES: every long-running job runs under a watcher that restarts it by pid and records the restart; the fleet lane reads the chain every 15 minutes (blocks per minute, key count, time since the last lock, seed slot counts, the home machines' peers); infrastructure faults are fixed as they occur under the pid rule (a dead seed, a banned IP, a stalled roll, a filled disk, a dead pod); only consensus-level intervention stays forbidden (no hand-made lock, no reorg by hand), and a consensus-level stall is the first line at 08:00 with its evidence; the rented ceiling stands. (2) MAXIMUM SPEED: Phase 1 of the execution plan starts tonight with the lanes that stay; new lanes are permitted overnight for a defined Phase 1 deliverable with a morning clock, each briefed with the four rules (pkill refused; fleet binaries only from the shipper's kit, "cut" only on the build-1 read-back; nothing published from the Mac but the site lane's script; a slip once, no "done" unread), the kill guard and the canary rule. Idle builders are a fault: build-5, 6, 8 and 9 carry work all night. +PHASE 1 (owner, clock UK): (a) the 2.0.2 cut with the fresh-install canary, the fleet roll and the three home machines mining on it: shipper (cut on green, about 23:15), fleet (the roll after, 01:00), relay (both PCs MINING-ON, PC 1 at the cap 04:50), the mini the shipper's; (b) the cross-backend same-work test on the frozen object across node, CPU, CUDA, Metal, OpenCL and pool: steward with the fleet (the spec page landed; the runs from the freeze, 06:00); (c) the two-node cache-history test: node with proving (03:00); (d) key succession H and W named from tonight's real next-pair run and the pin landed on the 2.0.2 line: node with proving (the RESULT 21:40; H and W proposed in the 08:00 read-back for the founder's word, the pin tool landed tonight, the real pin on his word); (e) the native 40/40/20 finality case and the backfill case on rented boxes: node (the plan 23:00, the runs on build-7/8/9 from 00:30, rows 07:00); (f) GOV-03's node build made reproducible: build-server (the build-id try 22:00, the result 23:00, the link-order try after if needed, 04:00); (g) the two-hour declared-load hold on 2.0.2 with the outcome ledger: fleet with proving (from the cut plus one hour, rows 04:00); (h) the mine-evict-prove-rebuild rows on 8, 12 and 16 GB cards: fleet with V6-07 (first rows 01:00, the 16 GB cell on the 2.0.2 modes 05:00); (i) the comparative rows and the KAWPOW chip model: the comparative lane (first landing 02:30, the chip model 09:00); (j) the live-dataset census on the frozen object: hash (overnight on build-6, rows 07:00); (k) the D5 rehearsal plan written for the morning: node (06:00); (l) /acceptance regenerated and /prize live: site (the regeneration's edge read 22:00, /prize 23:00); (m) PC 1's knee rows at 05:30 and the layer-8 verdict ready for 08:00: hash (the grid's cap 04:50, the knee rows 05:30, the verdict 07:30). +BUILDERS ALL NIGHT: build-5 the hash lane's P01 OpenCL/Arc reference generation and the live-dataset census's second half; build-6 the live-dataset census (hash); build-7 the enforced lane's succession and stage runs, then the finality cases (node); build-8 the finality cases and GOV-03's reproducible builds (node, build-server); build-9 the 2.0.2 chain and the two-node cache test (node); build-1 and build-2 the gates; build-4 the comparative lane's rows and the same-work runs (steward). +THE EVIDENCE CLASSES LANDED (21:31 UK, read back): box master 45b87ee8 (ci-evidence-added d7689a43, 89 checks, under the lock, pushed on try 1): rule 2 binding modified and deleted evidence files only (a new file landing free), the recorder refusing a missing or directory path at record time, the kills batch citing the two experiments' own files; F03 (1fd509d5) in its gate; rule 33's canary check, guards and registry case on ci-rule33-canary by 22:00. The 2.0.2 line 45e7b910: the miner suite green (33), the rest of its gate on build-9. (c)'s pair built, placement and canaries next. + +THE 21:3x TO 21:4x LINES. THE FIRST POST-ROLL LOCK: hub-1 checkpoint 620, block 8b2d4553, blue score 10,338, DAA 18,570, at 21:34:33 BST, signed 1,029 = 80.8 percent of active (63.2 percent of total, 63.3 percent of the table frozen at lock 619), finality active; 18,576 blocks, hub-1 peers 7, 4 keys over its last 300 blocks (the fleet's fixed miners); both public seeds listening with zero fleet inbound slots at 21:35 (the hub-only SEED set on 89 of 92 fleet boxes, applying at each 2.0.2 restart); the fleet's 15-minute watch from 21:34 (blocks per minute, the key count, the time since the last tallied lock, the five dial ports, the seeds' fleet slots, the mini's node, the fleet jobs' pids, hub-1's disk; faults fixed by pid with each restart logged; consensus never touched); the pid-only tree on master at 94ef14e7 (its four gate reds fixed inside 25 minutes; the founder's name scrubbed from box-dn3.sh by the pre-public scrub). A SECOND FRESH-INSTALL CLASS (the node lane, from build-1's two fresh-sync runs at 21:20 and 21:24 to 21:29): a fresh 7cfa422a node from genesis does NOT reach the tip: IBD takes the first 3,600 blocks, the executor reaches tip 1,943 (the last chain block of epoch 0) and stops; the IBD waits on "a class v5 header chunk waits for this node's execution state (class v5 needs the execution state after the epoch's seed block 92137840)" and after 300 s ends "this node's executor did not reach the class v5 epoch's seed block within 300 s", disconnects, re-dials and repeats; no "not held yet", no ban; so PC 2's 3,847 (the proof-missing class, the hubs' lost records) is the second stop, and this one comes first for every fresh node: a node fault of 2.0.1's own, the tip lag of 417c4a57 (the follower keeping the last two chain blocks of every path for the next pass) withholding the epoch's seed block at a chunk boundary while the class v5 catch-up waits for exactly that block: a deadlock at the first epoch boundary of every fresh install (the 17:58 fresh-join read passed because the seed was still in epoch 0); the fix: the lag applies only at a live tip (a sink whose timestamp is within ten minutes of now), on a stale sink the follower takes the path whole, with a known-failed test; 6c7abacb on successor-2.0.1 with the record-store take (a9ff0a25), the fresh-sync read on build-1 against the pair as the pass line (the node reaching the seed's tip from genesis through the epoch boundaries and the 3,685 to 3,847 band under lock 270); stage C's fresh joiners stalling at 3,600 the same class. PC 2's probe verdict (21:28:11 to 21:42): the empty-datadir restart's fresh 2.0.1 node stalling where the peerless boxes do (IBD 3598 blocks at 21:33:37, then "node behind", 0 MH/s); PC 2 mines again only on the 2.0.2 node entry (through its own update path the minute it publishes); PC 2's relay agent silent since 21:27 as it ran that restart, its revive through the signed-jobs channel by pid; the relay lane's two re-arm loops under a watchdog. The shipper: the app lane released (its quit fix ab37532c in release-2.0.2 at 7cb51f73); rules 17 and 18 on master (5fc4283d); hub-1's miner rolling on the 2.0.1 kit; the three rule-33 canaries briefed (the fleet kit on lp-4090-11, Windows on PC 2, the Mac DMG on the mini) with the steward's record form and build-1:/srv/canary//; the 2.0.2 node sha = 6c7abacb + the 2.0.2 bump about 22:50 (45e7b910 if red), the cut about 23:15, the entries after their canaries about 00:00. The 2.0.2 line 45e7b910 six suites green (33/95/183/143/38/19), the pair placed on both boxes, the canary be5f4068 on both empty nodes, chain id 0x1171: green end to end; 6c7abacb's gate on build-9. LANE D's 6.11 AND 6.12 ON MASTER at 885327402 (21:39, read back; the logs, the harness diffs, the batch family-gate-20261008b at manifest ac86d7910, method native; the pairing read back at 1a938abe4, both pack ids equal); lane D released. THE CENSUS LANE's LAST LINE (21:36): the (c''') read rerun at the final sha 1a938abe4 reproducing hl-v6-all 0x4de7b836cc40a4ea and reading identical to the b24dfc162 rows (harness class-v6-census-all4 c7e41d5f2); the freeze's census rows standing. FLOOR LANE 2's PLACED 64-REGISTER WINDOW CORE (21:32, read back from the pod; D3's acceptance-rule chip side; the genesis families, the register file clock-gated, 8 lanes; placed and routed on ASAP7, 563,339 cells, parasitics from global routing, the SPEF lost to the full disk; gate-level VCD, steady state from 150 and 600 run cycles; a model, never a lower bound): 9.45 pJ per lane-op at ASAP7 plus or minus 15 percent (+52 percent over its synthesis, inside the band); 6.6 at N5, 4.8 at N3, 3.4 at N2; k at the lock 1.07 / 0.77 / 0.55; the GDDR7 board at the 5090's lock 2.0x node for node, 2.45x a node ahead, 2.9x two ahead on the board identity; the window's residual against the placed gated base (6.7) +2.75 pJ per lane-op, +0.23 of k at the lock: THE WINDOW TAKES THE BOARD FROM 2.4x TO 2.0x node for node and 2.8x to 2.45x a node ahead for at most 5 percent per load on the card, a real defence, the expected 2.5x and 2.1x to the digit; F05: the chip's energy per hash at N3 from 0.82 to 0.95 microjoules with the window, floor lane 3's prefix-tree share 2.6 to about 2.2 percent; 86e5b0fb2 on class-v6-floor-k, gating with the landing hand; the crossbar, tile, core8r64 and core32 rows at 22:30. THE ADVERSARY DELTA LANDED (21:42): box master 01ce4e451 (Merge class-v6-adversary-docs-3 ac0bc17e; 89 checks): the document whole, the two cells and the batch (nine cases, ADV-05 included), no repeat of the steward's rows; thirty-two landings; the placed 32-lane full core's slip once: 22:45 to 23:15 (the slower Xeon host's run dying at the clock-tree step, the timing repair peaking at 252 GB on a 251 GB host; the fleet host's run in its global placement's last passes at 503 GB, the CTS repair skipped). THE WORKERS PAGE READ IN A PLAYWRIGHT RENDER on build-2 at 21:34:33 and 21:37:52 (the build-server lane): ten server cards (the nine boxes with their OS lines and "igneum-mini, the Mac build box, M6" with "no report yet"), the crew row (the nine boxes, MacBook-Pro, PC 1, PC 2), no hardcoded feed map (the FEEDS check false), the root 302 to /workers.html, the home page's six icon links and the five icon files served (favicon.ico 200); committed e344886c and 4b4d91d3 on build-server; the old dl bookmark's redirect stub queued on the serialized dl deploy; build-3 rendering up because its feed answers tonight (the box reachable again). THE ENFORCED LANE's SLIP, ONCE: the succession RESULT and batch 21:40 to 22:10 (pausing the attacker's miner made it worse, A never adopting the honest tip in 300 s twice; reverted to a running miner with a 600 s rejoin, 9c93fd95b, from 21:34 on build-8; the node's behaviour right on every carried record in every run, the open item the harness reading all six); its Phase 1 defaults: (c) the cache-history fast-time case on 3f672661 (a proof refused for context at carrier n then paid under a matching statement; invalid bytes cached once and refused at every later carrier without a second verify) on build-9 by 03:00; (d) W one epoch of the live object (pow_epoch_blocks DAA), H the first epoch boundary after a clean 24-hour run of the live chain on the 2.0.2 line with no pause and no reorg over depth 3, the fast-time runs as the evidence; (g) the outcome ledger over every fleet box's prover-state.json and prover.log at cut plus three hours, rows by 04:00. The pool seat: pool-2.0 986252e7 (the binding switch) read back; batch 03 landing. The shipper lacks ids for an Ember lane and a comparative lane (none spawned through the coordinator); F11 NOT RUN, dispatched, if no Ember lane exists. +THE FINALITY NATIVE-RUNS PLAN ON MASTER (21:4x, the node lane): 9e9d2532e (gate green on a470362a, read back, five mirrors): docs/plans/finality-native-runs-2026-10-09.md, PLANNED, the runs from 00:30: section 1 the 40/40/20 case past the window (honest; the equivocator on one island; on both, the 6.5 bound with two recovery certificates and lockKind "recovery" on both, the FAIL line a recovery lock presented as final; the pause-only variant); 2 the pause-only alternative with the historical-checkpoint backfill, a node with its finality state removed, every old key returning; 3 authority succession and strip, a voter's restart, certificates in reverse order (the hub-1 263-then-262 read), the seed boundary inside a pause; 4 the Sepolia verifier on a final and on a recovery certificate; 5 V6-09 on the minimum validator with the enforced-proving lane; 6 the two-node cache-history test (F01) on build-7 by 03:00; build-8 sections 1 and 2 (rows by 05:00), build-9 sections 3 and 4 after the 2.0.2 line's gate (rows by 06:00) and section 5 (rows by 12:00), build-7 section 6; every run under the lease pool with a pid file and a watcher; the rows into sim/results_v2.md with the registry batch by 07:00; the runs' lane spawned once the 2.0.2 sha is out (about 22:35), the harness extensions (the third island, the equivocating key, the backfill joiner) its first commits. + +A FREEZE-LEVEL FACT (21:5x BST, the steward's same-work hand, read back by the node lane): the frozen pack hl-v6-all (0x4de7b836cc40a4ea) was drawn with the genesis epoch seed edc4fa84 over a node1 state stream whose block is af89be5d, and IgneumEngine::epoch_for's class v5 pairing check refuses that pair ("the provider's stream is for chain block ..., not the seed block"); on the chain's own seed af89be5d the engine draws 2a1d6caab4c24564, not the pack's; the node and pool readers cannot re-check the frozen object AS PACKED, the CPU and CUDA readers can (no pairing check); the steward's ruling for the same-work test: a second job context on the chain-seed object 2a1d6caab4c24564 run on every reader first (the same-work row), the frozen-pack context a CPU/CUDA-only row with the reason. THE COORDINATOR'S DEFAULT unless the hash lane rules otherwise with evidence by 22:30: the D1 object is the generator tree 1a938abe4 with its freeze fingerprint and the dataset policy (the chain drawing per epoch from its own seed); the id named in F0's signing block and the D1 record the chain-seed draw read back equal from the node's engine and the CPU and CUDA readers on the same (seed, state) pair; the pack's id the research reference at its own pairing; the census hand and lane D resumed for one re-read each on the chain-seed draw (by 23:00 and 23:15); the shipper's 23:30 condition the engine's draw at devnet-4's epoch 0 seed from the placed pair equal to the CPU reader's; a FAIL or a disagreement reads "frozen pending" with this cause. THE CLASS: a pack is a (generator, seed, state) triple and the state must be the seed block's; the export tool paired them without the engine's check; the research reads at the mismatched pair are reads of the generator's draw at that pair, named so. +THE PAIRING READ BACK (21:46:55 BST, the node lane, from the placed (c) pair /srv/artefacts/200-e8773ff5/node-lane/igneum-miner, sha 2ee4893d, the 1a938abe4 tree): `igneum-miner program-id --epoch-hex af89be5d --era-hex af89be5d --day 20730 --class v6 --state ` printing "program-id class v6 attempt 0 id 442a1691b3e3507f seed af89be5d era af89be5d day 20730 state 1c583d35", EQUAL to the freeze CLI's export of the same class on the same inputs at 1a938abe4 on build-5 (generator 6, attempt 0, id 0x442a1691b3e3507f, loads 256, era label 93a14ac6; the hash lane's decision (a)): the D1 object is the tree 1a938abe4 (fingerprint 5f4d6dc6199294db89042171004e6420c1e5791e716d4b39b73d375818c10b6f, the class-v6-review freeze line) with the dataset policy; the id in F0's signing block and the D1 record the chain-seed draw 442a1691b3e3507f at (seed af89be5d, era af89be5d, day 20730, state abb58003), read back equal from the node's engine and the freeze CLI on one pair; the pack's 0x4de7b836cc40a4ea the research reference at its own pairing (seed edc4fa84 over the same state, refused by the engine's class v5 rule by construction); the steward's 2a1d6caab4c24564 the engine's draw on MIXED inputs (epoch af89be5d, era edc4fa84), a third instance; the same-work test's node and pool contexts on the real pairing expecting 442a1691b3e3507f; the chain-seed pack for the census and lane D re-reads /srv/artefacts/packs/hl-v6-all-chainseed.tgz on build-1; (c)'s canary condition PASS on this read. A 2.0.2-BLOCKING FAULT (the V6-07 sub-lane, 21:5x, with evidence): master's proving host after the V6-10 commit 7604fcf5d writes every guest input with the first field format = 3 (core/src/shard.rs GUEST_INPUT_FORMAT, check_input_format in the guest) while master's pinned guests are the 5 October pair (elf/igneum-prove-program.elf last changed at 15bb6cdd4, the manifest pinned_at 2026-10-05T16:20:38Z with no input-format or source-commit field; pin C held on p22-stage-2-pin); measured: a host built from box master cef5234b5 with cuda (66662219a3630772) executing fees-v1-shards2 shard 0 to "public values are 0 bytes, expected 328" on an RTX 3060 and an RTX 4060, the served 0317 host (71bc2438) proving and verifying the same shard in 14.1 s on the same card; nothing built from master's proving crate proves against the pinned guest; the fix by default: the host writing the input format the pinned manifest names (format 3 only when the manifest's pair is pin C, the manifest carrying input_format and source_commit), a known-failed test, and a CI check (a GUEST_INPUT_FORMAT change without an elf/ and manifest change fails), the enforced lane's by 22:30, the 2.0.2 kit's prover held on it; the V6-07 rows' default path on the served 0317 host, the proving rows on master's host NOT RUN with the cause. THE 2.0.1 LEDGER BASELINE (the enforced lane's dry run on lp-4090-01 and lp-4090-02 over 10.2 hours, pulled 21:45): 160 claimed jobs, paid 2, active 32 (submitted and waiting), expired 98 (97 unpaid after a submit, 1 held past its deadline), cancelled 28 (17 shards refused, 8 segment records refused, 3 chain failures), abandoned 0, the conservation holding; paid completions 2 segments and 47 shards, 3.72 IGN, median end to end 212 s, median time to pay 232 s; missed deadlines a median 24 DAA past with a median margin of 434 DAA at the claim; 11,862 of 13,489 GPU seconds wasted, 8,700 on records accepted and never carried in time; 0.2 segments an hour per two boxes, 2.7 percent of seconds paid; the loss the carrier side (97 of 160), not the prover (3 chain failures) nor the claim margin: the cell the 2.0.2 hold must move (the record store and relay caps, V6-08's lease and backpressure); if the unpaid share does not fall from 61 percent, 2.0.2 did not fix the carrier path; a 4090 box earning 1.86 IGN in ten hours at this rate, a home card nothing it can rely on; the 47-box pull at the cut plus three hours (about 02:15), rows by 04:00 per tier and kind with this as the 2.0.1 column. THE 17 GB CLASS CLOSED BY CONSTRUCTION (21:46, the build-server lane): two jobs deploys running at once again (pids 15981 and 27252, publish-jobs.sh from other lanes, the Mac's dozens of worktree copies carrying no lock); the fix at /opt/homebrew/bin/npx, a serializer taking an exclusive lock on ~/.igneum-guard/vercel-deploy.lock for any `npx ... vercel ... deploy`, inherited by the tree, released on exit, a second deploy waiting (the holder named) and refusing with exit 75 after 15 minutes, every other npx call passing through (the Homebrew link kept as a backup), tested with two concurrent fake deploys, logged; the script side on dl-publish-off-mac d0005844 (publish-workers.sh refusing on a Mac, publish-fleet.sh refusing unless IGNEUM_MAC_DEPLOY_OK=1 with a pid-recorded process group, publish-jobs.sh with the lock and IGNEUM_DL_PUBLISH=box, docs/release/dl-publish.md naming the split: the jobs triple movable to build-1's Caddy under a dl-jobs host with no app change after one last deploy whose vercel.json 307s the feed there); the flip held on the coordinator's word: after the 2.0.2 entries are live and PC 2 reads MINING-ON, on the shipper's word with PC 2's fetch through the redirect as the canary, else the morning's first item; the old bookmark redirecting (307) since 21:44:19. +MAIN'S RULING ON THE PAIRING (21:5x): the coordinator's default is the ruling (D1 freezes the generator tree 1a938abe4 with its fingerprint and the dataset policy; the chain draws per epoch from its own seed; the id in F0 and the D1 record the chain-seed draw read back equal from the node's engine and the CPU and CUDA readers on the same (seed, state) pair; the mismatched pack's id the research reference at its pairing, labelled; the census hand and lane D rerunning on the chain-seed draw by 23:00 and 23:15; the shipper's 23:30 condition the engine's draw at devnet-4's epoch 0 seed from the placed pair; a FAIL or a disagreement "frozen pending" with this cause); the class in the record; the export tool gains the engine's pairing check before any pack is named again (the hash lane, with the freeze note); the hash lane's 22:30 window for a contrary ruling with evidence standing. GOV-03's CLASS NAMED (21:49, the build-server lane): the cross-box igneumd difference a C __DATE__/__TIME__ compiled into a C dependency (mimalloc's version banner), not Rust nondeterminism: build-1's igneumd carrying "Oct 7 2026" and "02:53:46" (its sccache serving yesterday's object), build-8's "Oct 8 2026" and "18:58:34", build-9's "20:31:38"; .text and .rodata differing only by that string's shift, .eh_frame, .data and .data.rel.ro identical, no build-id note in any (zig linking none, so --build-id=none a no-op); the earlier "identical string tables" read (strings -n 12) missing the 11-character date, corrected once; the fix SOURCE_DATE_EPOCH set to the node commit's time for every ship build (zig's clang pinning __DATE__/__TIME__ to it), forwarded to the box and part of the object hash, folded into the ship-v3-sysroot change in tools/build-remote.sh; the test the 7cfa422a seed pair built on build-8 and build-9 byte-identical, the row PASS on that read by 04:00; the miner pair already bit-identical; the 2.0.2 chains holding the served prover pair (host 71bc2438, export 263bf4ce) unless the enforced lane's format fix lands green before the cut. +THE HASH LANE's RULING (22:00): the pairing holds; the freeze CLI at 1a938abe4 on the engine's inputs (epoch seed af89be5d, era 0:edc4fa84, day 20730, the node1 state 1c583d35 at block 159357) drawing generator 6, attempt 0, id 0x2a1d6caab4c24564, equal by its read to the engine's program-id line on the placed (c) pair; the pack's pairing (the genesis string seed edc4fa84 over the state after af89be5d) a deliberate research pairing from the W = 8 lane's 14:00 recipe, not a tool fault (the CLI having no seed-block check; the engine's check the chain's rule); the D1 object agreed as the tree 1a938abe4 with its fingerprint and the dataset policy; the five packs re-exported on the chain pairing as a parallel set (hl-v6-all-cs and partners) by 22:30. A DISAGREEMENT on the signing id (22:0x): the node lane's 442a1691b3e3507f (--era-hex af89be5d) against the hash lane's 2a1d6caab4c24564 (era 0:edc4fa84), the era field the question; closed by a live read: the node lane reading the program-id line the live devnet-4 node prints (build-1's seed log or lp-4090-11's miner log, epoch 0, day 20730, the era label) by 22:15, that id F0's; the census and lane D holding ten minutes for the inputs; the export tool's pairing check still owed by main's word (a CLI without the seed-block check having produced the mismatched pack). +THE SIGNING ID CORRECTED (22:05, the hash lane): the id read back EQUAL by the node's engine and the CLI is 2a1d6caab4c24564: the node lane's engine line at 21:46:55 reading "program-id class v6 attempt 0 id 2a1d6caab4c24564 seed af89be5d era edc4fa84 day 20730 state 1c583d35" (the era the chain's era seed edc4fa84, not the block hash), the CLI at 1a938abe4 on those inputs drawing the same (build-5, 21:48); 442a1691b3e3507f the CLI's draw with the block hash as the era (the hash lane's first read at 21:45:59, the census hand's 21:48 read the same wrong era), matching no engine run; F0 and D1 naming 2a1d6caab4c24564 at seed af89be5d, era edc4fa84, day 20730, state abb58003 (root 1c583d35), the node lane's live log line by 22:15 as the confirmation; the export tool's pairing check with its known-failed test on class-v6 tonight. THE PROVER DECISION (22:0x, the enforced lane, said once to the shipper): the host-format fault's root older than the format tag (master's proving crate writing inputs the 5 October guests cannot read since the D4 flag landed on master without its pin, 34f5914d2, compounded by stages 1 and 2 and the format tag; the only host proving against the served pair one built at the pin's source 15bb6cdd4, the 0317 host); the 2.0.2 kits ship the served 0317 prover pair (host 71bc2438, export 263bf4ce); a host rewritten in forty minutes and rolled to 47 boxes the fault class the founder forbade; the layout-by-manifest host on 2.0.3 with a card-measured proof first; on master before the cut by 22:25: the manifest's provenance for the served pair (source_commit 15bb6cdd4, guest_input_format 1), the host refusing at start when its input format is not the manifest's (naming the source commit to build from, never "0 bytes"), the known-failed test, the CI check (a GUEST_INPUT_FORMAT differing from the manifest's red unless the same diff moves elf/; the kit's prover built from the manifest's source_commit, never master's tip, until pin C moves). V6-07's floor patch 8c1fb754c (sha 9098c3e5, the one the landed server db37c38b was built from). +The coordinator's 442a line withdrawn once (22:1x; it carried the node lane's mis-quoted argument); F0 names 2a1d6caab4c24564 at (epoch seed af89be5d, era 0:edc4fa84, day 20730, the node1 state at block 159357, root 1c583d35, sha abb58003), the live node's program-id line the confirmation by 22:15; the steward's same-work hand's second context on it with hl-v6-all-cs.tgz and its CPU reference by 23:00. The pool seat released (22:0x): pool-2.0 986252e7 (the daemon refusing jobs while its node reads unsynced, blocked or reexecuting), batch 03 with the corrected cause and docs/analysis/pool/pool-pair-2026-10-08.md on master at 499c5c9b (21:49; UX-05 the vote-key PASS in the crate, the live pair NOT RUN; UX-04 NOT RUN with the synced-node dependency). The whole fleet's miners on 2.0.1 (hub-1's rolled at 21:43, 102 blocks in 343 s, zero refusals; the shipper's read). +THE SIGNING ID CLOSED (the node lane's verbatim lines from the placed (c) pair, igneum-miner sha 2ee4893d, the 1a938abe4 tree, on build-1): (1) 21:41:51 BST on epoch af89be5d, era edc4fa84, day 20730, the node1 IGSD1 state (sha abb58003): "program-id class v6 attempt 0 id 2a1d6caab4c24564 seed af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 era edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 day 20730 state 1c583d352bb9c75a06dadb8d82d42836ebe8afa82d0b413be87bf921741f1526"; (2) 21:46:55 with the block hash as the era: id 442a1691b3e3507f, the wrong argument for F0 (the chain's era seed is edc4fa84, the era VDF's output, not the block hash), named once; F0 AND D1 NAME 2a1d6caab4c24564 at (seed af89be5d, era edc4fa84, day 20730, state 1c583d35), the pack's 0x4de7b836cc40a4ea the research reference. (3) The live node's own line: the live chain is class v5, generator 5 (no class v6 epoch on any live chain, by construction of the freeze); build-1's seed and the fixed miner at the live epoch 2: "epoch seed 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 day 20734 (daa 8565): program and 256 MiB cache ready in 2604 ms; class v5 program id 81fbfa3aa413173f", and the (c) pair at 22:50:21 on those inputs with the live epoch-2 stream (igneum_getPowStateLeaves on 3c3fab43, 21,132 bytes) reading "program-id class v5 attempt 0 id 81fbfa3aa413173f seed 3c3fab43 era 0000...0000 day 20734" and the same id with era 3c3fab43 (class v5's id not reading the era, which enters from class v3's era rungs under the ladder, never active on devnet-4): the live node's id reproduced exactly by the placed (c) pair; the pairing witness for class v6 is line (1); for devnet-4's epoch 0 instance of the v6 object the (c) pair drew attempt 1 id a3764f0fc4c3d9f3 (era = genesis) and 1c54a42911b93e1e (era = zero) on the genesis stream at 21:48:17, the hash lane naming which era devnet-4's epoch 0 carries if the register wants that row. (The node lane's "22:50:21" stamp is its box's, two hours ahead of the Mac's; the Mac read 21:5x.) +F03 LANDED (21:52 UK, read back): box master ee3e0245 (ci-f03-manifest 69b51369, 89 checks, under the lock, try 1): release-manifest-check with the provenance rule, build-from-manifest, the same-work spec page, p01-vectors --job-context/--phase, the cells harness:release-manifest and harness:same-work, GOV-01 moved with the F0 page, the generators keeping evidence_records on regeneration; one fault read back: R2-F03-R01 reading NOT RUN though its cell recorded PASS (the normalizer's lifted legacy "record" entry, a reason with no run, counted by the combined decision), the fix (a reason-only entry never counting, dropped when a real cell lands) with the guest-format landing by 22:10; rule 33's gate on 5fc267f9. +THE EXPORT TOOL's PAIRING CHECK (21:58 BST, read back): class-v6 ce6e6902f45f40e727726cf3e0dbf0e1c924658a (the suite on build-7 151 passed, 0 failed, 8 ignored: the freeze's 147 plus 4): StateStream::check_pairing first on every --state command (export, bench, hash, hash-bound); a seed whose block is not the stream's block exits 2 with the engine's reason and writes nothing; a string seed never pairs; --unpaired-state recording a research pairing on purpose and printing it; the known-failed test (tests/pairing.rs) exporting the hl-v6-all shape and reading the refusal, then the same with --unpaired-state and the chain's own pairing; the generator untouched (state.rs, main.rs and the test only), the frozen object still 1a938abe4's draw; the five chain-paired packs exported at 1a938abe4 on build-5 (hl-v6-all-cs 0x2a1d6caab4c24564; the four partners' ids and shas about 22:15). +RULE 33 LANDED (22:01 UK, read back): box master d2e1c192 (ci-rule33-canary 8687874e, 91 checks, under the lock, try 1): tools/ci/canary-check.sh (the record tools/ci/canary/.json, one file per sha with an artefacts list fleet/windows/mac/hive, each block its own non-AVX-512 box and eight read-back lines; --artefact ; --form), the guards in packaging/ota/publish-manifest.sh (--release-sha, refused without a PASS block for each platform published; loopback test writes and --verify-only not gated) and publish-public.sh (the sha from the manifest's "release: " note; hive its own block), deploy-win.sh covered through them; the registry: INT-07 BLOCKED on cell canary:fresh-install (no 2.0.x sha with a record; the shipper's 2.0.2 canary the first), POW-03 and POW-07 citing their own files (the merge replaying modified batches too); the guest-format check (a format bump without elf/ red at the merge; a manifest field off the source red only without provenance) and the recorder's lifted-record fix queued behind it. +THE CHAIN-PAIRED PACKS READ BACK (22:03 BST, build-1 /srv/artefacts/packs/, sha256sum -c OK): hl-v6-all-cs 0x2a1d6caab4c24564 attempt 0 loads 256, tgz a4742bfefdc632a1f3977c9dfb8f66417c2f132436dcd9ad72c32a608638d6ca; hl-v6-all-nowin-cs 0x4e1897e6ee262228 attempt 0, 6ef02eeea5d078f5215b09145239e27a83c8a41e6bccb11bd638d3170b0538bb; hl-v6-foldrw-cs 0xb53d00f2bf629076 attempt 0 loads 128, 4f2abd36aee600d97d230b5b09f10680f228448007c6c7062fca737bf6351748; hl-v5-nowin-cs 0xa02b1a9dee6b8587 attempt 1 loads 128, e2eb2840879d264a8d3562751f983ce57d821087ae6d52796eccf3ef48484b8e; hl-v6-all-prefix-cs 0x2a1d6caab4c24564 (the F05 prefix text), a9e01f13d6f26f133504c6d62e73b86ee460b53d8798218100b2f250e0ac8ca5; all at 1a938abe4 on the pair epoch af89be5d, era edc4fa84, day 20730, state abb58003, generator 6, each with identity.json. The P01 CPU reference of the signing object: /srv/artefacts/packs/p01-vectors/hl-v6-all-cs.txt, 1,000,000 lines, sha256 b77c61d874aed238394fd0e556113192acf21e0762871b1044796f3fbaea338a (first line "0 2394c9f10f9447a2", last "999999 cdaadae412c7d9c1"), build-5 21:55 to 22:02; the OpenCL read on PC 1's RX 7600 queued behind the knee rows; the Metal read the morning's; the row docs/analysis/class-v6/rows/pairing-20261008.md with batch hash-lane-20261008-batch3 landing next. +UX-01's WRITE-BACK ON MASTER (21:43 UK, read back by the relay lane): dc94dda9 (the batch ux-01-20261008-win-2.0.0.json replayed; manifest aa354ed5; method team-reported; the cell pc:install-update covering UX-01 as team-run evidence with the P10 study NOT RUN, UX-07's install classes and OPS-03's update path; the evidence docs/plans/evidence/UX-01-20261008.md with PC 2's 2.0.0 and 2.0.1 read-back lines and the two network classes); four earlier attempts refused by rule 26 while other lanes landed docs: the shape that lands is a branch carrying only the batch, the cell and the evidence, never the registry or the generated page (the class for every lane's write-back). +GOV-03 PASS (22:05 BST, the build-server lane, read back on the boxes): the 7cfa422a seed pair built on build-8 and on build-9 byte-identical (igneumd 6502b3c7d58a16dfa5adaef30ffc696714f1a09f117648c62b4973fe129a8075, igneum-miner b4748308e0887982c7f2ce3f06bc64802aa6ea9780adb61413fcbb8934f69e37), the C banner reading the commit's author time ("Oct 8 2026" "18:44:40"), the fingerprint cbc5bd0aa10585c8 in both; the exact cause under the date: SOURCE_DATE_EPOCH exported but sccache's server compiling C in its own environment so clang never saw it; the fix defining __DATE__/__TIME__ on the compiler command line (both caches keying on it), ship-v3-sysroot 5c494c08; one class left: a non-ship native build through sccache still baking the wall clock, only the --ship path pinned. THE V3 KIT (a correction first: the 2,018 zmm lines were blake3 1.8.3's AVX-512 assembly, cpuid-dispatched, not glibc; under zig glibc is dynamic): the certified build --ship seed/linux at target-cpu x86-64-v3 with blake3's pure feature (no AVX-512 code linked), -mpclmul for rocksdb's crc32c, an ISA gate on the box failing the run on any AVX-512 line; read back zmm 0, ymm 371,416, runs ("igneumd 2.0.1"), GLIBC_2.34 floor; the hive/rig class defaulting to x86-64 (v1) on purpose (common HiveOS rig CPUs without AVX2; a v3 igneum-miner dying with SIGILL), zero zmm there too; the branch pushed not merged, the shipper deciding if the 2.0.2 cut takes it (green before the cut); the cost blake3 pure dropping the AVX-512 hash path on AVX-512 seeds, hashing only; the 2.0.2 dual-chain script staged (app202/run-202.sh, asserting the served prover pair, a kit-isa line per binary); two gaps for the shipper: the DMG and the Setup exes outside the chain (the Mac lock, the PC job), the lab hive's archive renamed to the brief's name with its inner directory igneum-lab/. +THE SUCCESSION CASE PASSES (22:04 UK, the enforced lane; build-8, 21:34 to 22:04, the F01/F02 fix node 3f672661, floor 360, window 300, two real proofs per phase): below H the next pair refused on its pair and the prior on its statement; in the window both on their statements, none on a pair; after H+W the prior on its pair and the next on its statement; nothing paid; no stale refusal replayed; the batch enforced-proving-20261008-02 (harness:proving-enforcement, seven cases, manifest 3f672661) in the gate with the harness and docs/design/key-succession-schedule.md, THE H AND W PROPOSAL for the 08:00 read-back: W one live epoch (3,600 DAA); H the first epoch boundary at least an epoch after a clean 24-hour run of the live chain on 2.0.2 (no pause, no reorg over depth 3, every fleet node on the cut); pin C the next pair, the served pair the prior; the evidence table carrying the five runs. Phase 1 item (d)'s evidence complete. + +A CONSENSUS-LEVEL FAULT (22:08 BST, the node lane, read from build-1): igneum-devnet-4 split into at least four chains at the class v5 epoch cuts. The matrix (eth_getBlockByHash on each node at 22:06): build-1's seed (188.40.146.49:26631, a dial target) stuck at executed tip 6,162 (daa 10,799, its epoch-3 boundary) since 21:34, holding 1a87e870 as chain block 5,760 (its epoch-3 seed); the hand (26671) holding the same 1a87e870 at 5,760 then its own epoch-4 seed 550eab88 at 7,988, tip 8,144 (daa 14,474), still mined; the light reader (26882) holding neither, its epoch seed f8dffb78 at 5,560, tip 6,919 (daa 20,496), locking checkpoints 681 to 683 at 59.7 percent of total; the fleet's chain (peers 213.173.111.10, 203.57.40.129, hub 64.119.209.250) on 756ff5bd as its epoch-3 seed, absent from all three; the seed refusing every fleet epoch-3 header ("header rejected before the PoW engine: invalid proof of work", strikes, bans), the fleet banning the seed for 600 s (N6) for re-advertising its refused tip f09df69b; the seed's, the hand's and the light reader's IBD attempts against fleet peers dying in the 300 s wait "no published state stream after 756ff5bd (the block is not on its chain)"; the seed logging voter EQUIVOCATIONS at indexes 76 to 83, 250 and 262 and holding locks 260 and 261 by certificate with zero local votes: the voters split across the chains. THE CAUSE (two node classes, both in 2.0.1 and in 6c7abacb): (1) the class v5 IBD catch-up syncing bodies once, before the first deferred chunk, so every further epoch boundary costs the 300 s state wait and an IBD restart (measured on 6c7abacb's fresh node: 1,943 for 300 s, then 2,645 to 3,843 in 20 s after the restart); (2) structural: a node whose executor captured block X as epoch N's reference validating only epoch-N headers seeded by X; another chain's epoch-N headers (seed Y) refused for want of the state after Y, their post-cut weight never counting, GHOSTDAG never moving the virtual to Y's chain, the executor never capturing Y: once two nodes differ at a cut they can never rejoin, each extending its own chain with whatever miner points at it; a lock above the fork (the seed's 260/261) making it permanent: THE D5 PARTITION HAPPENING LIVE WITHOUT A PARTITION, THE EPOCH CUT ITSELF THE PARTITION. The fresh-sync read on 6c7abacb's pair red for the wrong reason (1,943 to 3,843 across the first two cuts, then 3,844 whose carried proof no build-1 node serves in 20 s, the records lost after the hub reorg; the lock rule unable to cover it on a fresh node; the seed's "tip" a dead fork). THE NODE LANE'S DEFAULTS unless main says otherwise by 22:40: (a) fix (1) on the 2.0.2 line as one commit by 22:45 (bodies re-synced to the highest validated deferred header on every state wait, the 300 s budget reset on progress), the six suites read back; the 2.0.2 name sliding once from 22:35 to 23:30; (b) fix (2), the foreign-seed capture (on the state wait for Y, the executor executing Y's selected-parent chain from the fork point in a scratch state and publishing the stream under Y's hash, so the headers validate, the weight counts and the virtual can move), designed with the enforced and hash lanes (class v5 design section 5) and built on the 2.0.2 line tonight, suites by 02:00, the two-node cache test's shape as its gate; (c) nothing by hand on the chain (no seed restart, its datadir the fork; no lock; no reorg); (d) the 23:20 freeze line standing on its own inputs (the D1 candidates gated on the live line before the split). Devnet-4 as it stands cannot heal with 2.0.1 binaries on the fleet; a reset (a devnet-5 genesis, or a coordinated restart from one chosen chain's snapshot) main's call, the node lane recommending a reset once (b) is read back; the coordinator's default to main: the chain left as it stands overnight under the guards, the 2.0.2 roll with fix (1) proceeding, the reset the 08:00 read-back's first item with (b)'s evidence, no genesis cut without main's word; the shipper's cut held on fix (1); the canary's pass line the node reaching the FLEET's chain tip from genesis through every cut, never build-1's seed; the fleet lane reading which chain the hubs, the rented miners and the voters' majority weight extend by 22:40. +MAIN'S RULING ON THE SPLIT (22:1x UK): (a), (b), (c), (d) standing as the node lane set them; the 2.0.2 cut proceeding with fix (1) for the fleet kit (the fleet one chain, the roll keeping it one); no home-machine entry (Mac, Windows, HiveOS) publishing until its rule 33 canary proves a fresh install reaches the fleet's majority tip (within ten blocks of hub-1's) with the entry's dial targets verified by the fleet lane as nodes on that chain; the entries waiting on fix (2) if it needs it (a fresh install syncing to a dead seed's chain worse than no entry); the reset the founder's decision: the chain left overnight under the guards, the fleet rolling 2.0.2 with fix (1), the home machines waiting, the 08:00 read-back opening with the split, its cause, fix (2)'s evidence and the node lane's recommendation (a devnet-5 genesis against a coordinated restart from the majority chain's snapshot) with the cost of each; no genesis cut without his word. FOR THE REGISTER (main's words): the epoch cut acted as a partition and the network did not heal on its own: the D5 property tested live and FAILED on 2.0.1; the fix is software rolled, never hands. + A SHARED-DEVNET FACT FROM THE FLEET (not this lane's, with the shipper and the infra lane): the Hetzner live seed 188.245.5.161:26611 is still on the old override object (digest eada4bda) 1 h 40 min after the 0.3.20 sweep (the fleet never touches Hetzner nodes, so it was outside the sweep); the 0.3.21 wipe canary c22-1 took five digest-mismatch rejects from it; an app with the packaged peers is refused at the seed and syncs through node1 and the hub only, a fresh joiner with only the seed cannot join, the 14 voters and the hub are unaffected; the owner puts the floor file ov16-floor-900000.json (sha 294f1f80) and the c4459193 pin on it. 0.3.21's STAGING (the node lane): the order dry-merges onto 55768f88 with nothing moving to 0.3.22; the late-join fix is 52e96c94 (70e4601e rebased onto 55768f88, exec suite 33 green with both new tests); f067f7c1, b0444f51 and 437f0438 merge clean in order; 2e32d5f6's one conflict (DST_ADDRESS beside pool-finish's DST_BINDING in consensus/core/src/finality.rs) kept both; the live-file digest eada4bda after each (every switch at never); the staging waits on the shipper's sweep-end word; the re-pin held. PC 2 DOWN AGAIN (main, 16:5x UK): the founder takes PC 2 down for cable work (PC 1 back but his desk); both PCs out of the sweep's waves, each updates on its poller on return; no PC job to PC 1; the Windows G1 completed before the outage, nothing reruns. 0.3.21's SECOND GATE LINE on 55768f88 (sha256 279b1b690e854fc9): the ten-minute mixed-version gate beside the 5899f603 pair, 13:37:40Z to 13:47:52Z, SUMMARY PASS (one digest b0afb2ee on five nodes; 223 new and 381 old blocks accepted by the old hub, 0 rejected; counts equal at 319, 486 and 604 through both clean joins and the restart step at 13:45:22Z; no panic); the node lane's two lines on 0.3.21's first candidate complete, in plan 6.9 on ca3-v4-node; the fleet's set on it (the bare-child 12 GB line, the wipe, the kept read, the cases) is the fleet's. 0.3.21's FIRST GATE LINE on 55768f88 (sha256 279b1b690e854fc9, the string read back; pairing igneum-pow 8c728ca3 at byte 5): the digest gate 13:35:41Z to 13:37:19Z SUMMARY PASS (a89be8a7 on both binaries with the peers; db9a85f9 refused, no peer; the live file's eada4bda unmoved); the ten-minute mixed-version gate from 13:37:40Z, line about 13:50Z. The 0.3.21 order as the shipper sent it: 55768f88; f067f7c1 and 70e4601e; b0444f51; 6eb21fc9; db28d331; then the re-pin from 8bdcbdd8 on the coordinator's word; suites between, the digest read after every one; the mirror's release-0.3.20-node back at the pin c4459193, release-0.3.21-node open at 55768f88. THE LATE-JOIN COMMIT (N9's second half, the node lane): 70e4601e on the box mirror as branch proof-hold-fix, from c4459193, two files (igneum/exec/src/proving.rs, protocol/flows/src/v10/proving.rs); the gap was the fetch side on the joiner (the served record ran the native check against the joiner's trailing exec state before anything was stored, the check refused it, the proof was never held, the body rule read "not held" for 20 s and failed the IBD); the fix holds the proof by hash before the checks (the pool entry still needs them) and the serve side says when it holds fewer than asked; the exec suite 32 passed at 13:26Z with the known-failed shape first, the flows check green 13:28Z, igneumd on build-1 at the 0321 worktree path built 13:32Z, sha256 17649eeb2f7d1290, string read back; with the testnet lane (the resume form, B alone); it joins the 0.3.21 staging as its own commit. THE WIPE CANARY ON c19-1, c4459193 (sha 45be9b02d1b002f5, string read back): FORM END rc 0 at 13:50:53Z. Wipe synced 13:35:50Z (57 minutes, inside the 98-minute class); mining 13:36:00Z to 13:47:07Z, 66 mined, 66 accepted, 0 rejected, isSynced true at the tip throughout; the hub holds 41 of its blocks in its last 700 with 0 rejects (13:47:09Z); the restart on its kept datadir at 13:47:15Z: the old process stopped at once (the new process's first lock line seven seconds after the marker; the watchdog held nothing, the b7cc37e7 fault closed), synced again at 13:48:39Z after 84 s, 109 templates read with max 3,432 ms and 0 timeouts; the kept read on pool-1's 0.3.17 copy on the same pod passed at 13:38Z (the rewrite line once, a clean second start). The pin's set on c4459193: the digest gate PASS, the mixed-version gate PASS, the wipe canary PASS, the kept read PASS, the restart PASS, the 12 GB line proves and verifies (paid is a race, not a gate); CASES END from c20-1 (about 14:50Z) is the last pin line. THE INTEROP FACT stands from the void run: the 5899f603 hub accepted 235 object-byte-5 blocks from the 8097d600 node with 0 rejected, one digest on all five nodes on the live sixteen-field file. The gates: the digest test and the kaspa-pow vector test (the amended devnet epoch-0 id 1a4230699a6b9c60 must equal, c120d7963abdcd96 must differ, the v3 control unchanged) on the box; the mixed-version Devnet 2 gate (the amended 0.3.20 node beside a 5899f603 node for ten minutes on the live file without the v4 fields) after the Mac build; the fresh-join canary the 0.3.20 cut's | | Main's rulings (7 October, morning) | no generator change to v4 on the live devnet; the record's null is the window model with numbers, sent by the hash lane to the attack-pass lane so AP-F8-1 re-gates against it; a fault beyond the model (a low-entropy source at site 15) stops at the coordinator with the two options priced (a 0.3.19 class amendment before the flip, or the flip held at the floor), nothing shipping without the founder's word; the tighter tail, an acceptance bound on the hot-set share, is a CLASS V5 item (sent to the v5 lane a6410f3b8abefb762 with the 64-seed census as its gate; the bound's number follows from the model) | diff --git a/docs/plans/igneum-2.0-register.md b/docs/plans/igneum-2.0-register.md index 2e7b1c80e..c790d3659 100644 --- a/docs/plans/igneum-2.0-register.md +++ b/docs/plans/igneum-2.0-register.md @@ -41,7 +41,7 @@ Built 8 October 2026, 18:3x BST (the founder's order: no feature left out, no st | 35 | D4. A five-year coexistence model (p. 13) | Profit-maximising operator simulation: mine, internal prove, external prove, off; under a proving demand spike, a token price fall, a major prover leaving, a specialised entrant in either market. Pass if pricing and capacity rules restore service without an administrator. | research lane (ad6a2bd47d4a46105) | 21:00 tonight | in flight: operator-simulation.md landed 859b3daa, the four shocks run on floor lane 3 by 20:15 | docs/analysis/class-v6/operator-simulation.md | INC-03, INC-04 | | 36 | D5. A no-rescue network exercise (p. 19) | Prerequisite: proof verification enforced in consensus (verifier_in_consensus, proof_rule_active_from) live on the exercise network. | enforced-proving lane (a6e8f84588b809d62) | landed | PASSED: verifier_in_consensus live on igneum-devnet-4 from block zero (4cdcc488), the seven refusals and the fast-time crossing PASS at 17:22 | docs/spec/proving-enforcement.md | ZKP-01, ZKP-08 | | 37 | D5. A no-rescue network exercise (p. 19) | No founder-operated mining, proving, aggregation or mandatory distribution infrastructure. | node lane (a283f5f0d364ceef0) | days two and three | open: the three ex-testnet seeds held for it | | OPS-01 | -| 38 | D5. A no-rescue network exercise (p. 19) | Cross epoch boundaries, interrupt signing, partition the network, remove major operators, hostile proof submissions, independently written clients. | finality lane (aca0f5ed924a2a99b) | days two and three | in flight: the partition row landed ec9ea053d (rule v4); the rest of the scenario table owed | docs/spec/finality-guarantees.md | FIN-02, FIN-03, FIN-04, FIN-08, ROT-01 | +| 38 | D5. A no-rescue network exercise (p. 19) | Cross epoch boundaries, interrupt signing, partition the network, remove major operators, hostile proof submissions, independently written clients. | finality lane (aca0f5ed924a2a99b) | days two and three | in flight: the partition row landed ec9ea053d (rule v4); the rest of the scenario table owed; LIVE, 8 October 22:08 UK: igneum-devnet-4 split into at least four chains at the class v5 epoch cuts (a node that captured block X as epoch N's reference validates only epoch-N headers seeded by X; another chain's headers are refused and their weight never counts, so the virtual never moves and the nodes never rejoin; a lock above the fork makes it permanent): the epoch cut acted as a partition and the network did not heal on its own, the D5 property tested live and FAILED on 2.0.1; the fix is software rolled (the IBD catch-up re-syncing bodies on every state wait on the 2.0.2 line by 22:45; the foreign-seed capture built tonight with suites by 02:00), never hands; the reset the founder's decision | docs/spec/finality-guarantees.md | FIN-02, FIN-03, FIN-04, FIN-08, ROT-01 | | 39 | D5. A no-rescue network exercise (p. 19) | A withholding concentrated prover: replacement operators, usable inputs, reassignment, explicit behaviour during proof delays. | enforced-proving lane (a6e8f84588b809d62) | days two and three | open | | CAP-07, INC-08 | | 40 | Pools, software and participation (launch requirements) (p. 14) | Vote keys stay with the miner at protocol level: the member's retained voting key committed into its work, payment aggregation separate, verifiable, pool identity substitution resisted. | pool design seat (a3832b1c3b274b310) | tomorrow 18:00 | in flight: the design on master 637e508b (the header commitment, the key on job and share); the code the pool branches rebase first | docs/design/pool-vote-key-commitment.md | UX-05, FIN-05 | | 41 | Pools, software and participation (launch requirements) (p. 14) | Non-custodial payouts, practical minimum payouts, local work verification, low-bandwidth participation (P2Pool as precedent, not code). | pool design seat (a3832b1c3b274b310) | tomorrow 18:00 | open: designed in the same file | | UX-04 |