Commit graph

893 commits

Author SHA1 Message Date
igneum-labs
ee64b8b8c0 Merge remote-tracking branch 'box/master' into site-2.0-followup 2026-10-08 20:12:49 +00:00
igneum-labs
21027145fd Merge remote-tracking branch 'box/master' into site-2.0-followup
# Conflicts:
#	docs/plans/igneum-2.0-test-registry.json
2026-10-08 20:12:26 +00:00
igneum-labs
71ea298e14 Merge v608-proving-tasks 01fa7cd2 into master (gate: green on 01fa7cd2, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 20:11:49 +00:00
igneum-labs
55d0830f73 Merge master 63e11d345 into ci-int-suite under the master-landing lock 2026-10-08 20:10:21 +00:00
igneum-labs
63e11d345d Merge ac2f34c9a ac2f34c9 into master (gate: green on ac2f34c9, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 20:07:46 +00:00
igneum-labs
f3c24e202f Merge floor-eco-record-02 c2fc307c into master (gate: green on c2fc307c, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 20:07:38 +00:00
igneum-labs
6898c72dbc Merge worker-select-pass ed93b542 into master (gate: green on ed93b542, recorded by tools/ci/pre-push.sh; landed on the box mirror under the exception declared by main: main's ruling, 7 Oct 2026 19:5x UK: the GitHub account is suspended, lanes land on the box mirror's master, the box gate stamp is the verdict; GitHub gets the fast-forward when it answers) 2026-10-08 20:07:26 +00:00
igneum-labs
654dc454a2 Merge key-succession-pin-tool 826c4335 into master (gate: green on 826c4335, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 20:06:05 +00:00
igneum-labs
d1ea621109 The 2.0.1 Windows entry on the index and the manifest (2.0.1 on every platform; 2.0.0 withdrawn, the line under the release note), OPS-03's row moved with it through the recorder
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 20:04:01 +00:00
igneum-labs
521d75680b P01 part A, the CUDA half complete: four packs on three cards, twelve runs, zero disagreements
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 20:03:26 +00:00
igneum-labs
c2fc307c5d ECO-05 batch 02 on the RX 7600's class v6 cell (15.79 MH/s measured, the knee 5.4 modelled; the metered knee is batch 03) and the B580's class v6 cell (11.0 measured, watts estimated): one more sustainable world (10R at 0.10, by 1.6 percent on the 7600's entry), 0 of 108 cells pass the envelope, RUN FAIL unchanged; the driver's override hook carried; the cubes and the registry rows recorded at manifest efd03e2c
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 20:00:11 +00:00
igneum-labs
156f3f54eb Merge hash-registry-1 ac89897c into master (gate: green on ac89897c, recorded by tools/ci/pre-push.sh; landed on the build mirror) 2026-10-08 19:59:34 +00:00
igneum-labs
12a188296e Merge pool-review-b-docs 9511e78b into master (gate: green on 9511e78b, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 19:58:33 +00:00
igneum-labs
ac2f34c9af The 2.0.1 Mac and HiveOS entries on the index and the manifest with OPS-03's row through the recorder (site:manifest, NOT RUN in progress); the reference population and the cost per accepted unit of work served on the litepaper and the facts page (floor lane 3's table on master at a954c2b7); the box master merged
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:57:25 +00:00
igneum-labs
09eff1c3dd Test map: the site:manifest cell (OPS-03, partial coverage: the manifest regenerates from its sources; the separation is the node suites') so a manifest regeneration moves its row through the recorder
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:55:44 +00:00
igneum-labs
01fa7cd2bc V6-08 build, the fleet side: box-prover.py posts each claim as a lease (igneum_claimSegment) and skips a refusal, the chain-health backpressure gate (blocked, re-executing, discontinuous, lagging, stalled, finality paused, unreachable), TASK=segment|shard|auto with one-shard tasks under the segment line and MAX_SHARDS, the preflight (host and export sha256 against the shipped manifest, --mode id against the node's pinned pair, ldd, nvidia-smi, exit 4 naming the file), the fourth outcome abandoned (lease, restart) and the flow line; prover-outcomes.py reports abandoned, the kind column and the conservation identity per kind; both self-tests known-failed first, the gate runs box-prover's
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:55:43 +00:00
igneum-labs
a954c2b7d1 Merge counter-asic-4-docs-3 6e9e532a into master (gate: green on 6e9e532a, recorded by tools/ci/pre-push.sh; landed on the box mirror under the exception declared by main: main's ruling, 7 Oct 2026 19:5x UK: the GitHub account is suspended, lanes land on the box mirror's master, the box gate stamp is the verdict; GitHub gets the fast-forward when it answers) 2026-10-08 19:55:03 +00:00
igneum-labs
ed93b5426f F10 rows, I05 epoch tag, I06 pack geometry rule across the three hosts
Rows: docs/analysis/review-2026-10-08-b/f10/rows.md with the pod rows (5090: steady
70.4 MH/s select against 63.7 full, 86.2 against 88.7 J per 2^24 batch; 3090 at the
card's clock 30.1 against 27.6, 184.6 against 197.9 J; the 3090 pod's power cap at
210 MHz named per run; ptxas 92 and 87 registers, no spills) beside the kernel timer,
the emulation and loader logs, the mini's geometry and serve logs. serve-bench.sh reads
a kit pack's seeds from program.h, execs the worker in its coproc (the recorded pid
and the CPU seconds are the worker's).

I05: every worker's found line carries a fifth field, the epoch the hash was computed
under (16 hex): CUDA and OpenCL from the pair, Metal from the job; select-check.sh and
serve-check.sh assert it, the swap case with pack B's epoch. The miner reads four or
more fields.

I06 (gate INT-10): one normative geometry rule. packfile.h reads the word count at
64 bits and refuses a count above 2^32 - 1 and a log2 that is not the floor of log2 of
the count; proto-metal/main.swift gains packGeometry with the same rules and refusal
words, sizes the day's dataset from the exact word count before any Metal work, and
--geometry-check <vectors> [pack dir...]. emu/geometry-vectors.txt is the shared
conformance file (ds55, ds85, ds115, the power-of-two sizes, wrong count or log2,
overflowing counts, missing leaves, leaves on a v2 pack): 19 vectors green in C on
build-7 (packfile-test.sh) and in Swift on the mini, with the two checked-in packs.

Record: cells harness:cuda-select-check (ROT-01, POW-01 partial) and
harness:pack-geometry (POW-02 partial); batch f10-worker-20261008.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:53:24 +00:00
igneum-labs
9511e78b15 Pool review B record (documents and the registry batch only): docs/plans/pool.md section 11 (F12, F13, INT-04, INT-15 closed by construction on pool-review-b 786e09cc: what was, what is, the regressions, the consequences per tier), spec 09 section 9.3 (the frame bound while reading; binding v2, the exact bytes; welcome, authorize, authorized fields; the job row carries master's vote_key_hash and the pool line's seed fields together), the test map's suite:pool cell and the batch pool-review-b-20261008-01 (51 of 51 on build-6 against successor-2.0.1 515ba674, cargo check green against the manifest's node 7cfa422a; evidence build-1:/srv/artefacts/tas/pool-review-b-20261008-01/). The code lands on the release tree (the coordinator's ruling, 20:5x UK): the pool line sits 581 commits off master with conflicts in app, workflows and tools, and master's igneum-pow fingerprint a65e4c5a differs from the freeze's cbc5bd0a, so the pool crate builds from release-2.0.1 or the pool line, never from master's own igneum-pow (F03)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:52:09 +00:00
igneum-labs
ac89897c6a test registry: the hash lane's batch 2 (the fleet's F05 rows on the post-review object as bench:fleet-pods evidence; suite:pow on the review-fix tree)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:51:34 +00:00
igneum-labs
6e9e532a58 Test registry: floor lane 3's batch eco05-20261008-01b (ECO-05 FAIL, ECO-01 and ECO-08 RUNNING at manifest 24883757; the rows move with their evidence files carried by this landing; the same run as batch 01, no result changed), recorded on the lane's word
The registry's evidence rule 2: eco-05-results.md and sim/economy/coexist/README.md change in this landing, so the rows that name
them as evidence move in the same diff, by their owner's batch through tools/ci/test-record.mjs.
2026-10-08 19:47:37 +00:00
igneum-labs
12c8fa4861 Registry: census:class-v6 PASS (run census-v6-20261008-2034: the freeze's object hl-v6-all 4de7b836cc40a4ea at class-v6 9415e9c89 through the sub-version 3 harness; evidence docs/analysis/class-v6/census-packs.md sections 0 to 7)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:47:22 +00:00
igneum-labs
da5bc75fb8 Merge the box master before the relanding (the download buttons from the refreshed index: the 2.0.1 Mac and HiveOS entries; the manifest's versions with them)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:47:14 +00:00
igneum-labs
cff719a04b Merge box/master 27a8f4bbd into counter-asic-4-docs-3 (rule 26 again: master's map, page and registry taken, the lane's cell re-added, the batch re-recorded) 2026-10-08 19:46:35 +00:00
igneum-labs
27a8f4bbdf Merge reference-apps-f04 1ca1eed7 into master (gate: green on 1ca1eed7, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 19:46:20 +00:00
igneum-labs
f8df3d8d26 The master-landing lock: one holder on build-1 for the whole landing, master merged into the branch under it, the merge stamped as the union of two gated parents
merge-to-master.sh takes /srv/builds/_locks/master-landing on build-1 (an atomic mkdir with the holder's pid, host, branch and start; waiters poll every 10 s in order up to a 20-minute cap; a holder older than the cap is reaped; released on exit, a kill by pid file included) for the whole landing. Under the lock, when master is ahead of the branch, the script merges master into the branch itself (the registry rebuilt from master's copy with the branch's batches and notes replayed, the map merged structurally, the page regenerated, the suites regenerated) and stamps that merge as the union of the branch's green stamp and master's own landing, so the push takes the light gate; a landing can no longer be refused for a master that moved, and lanes wait in the queue instead of racing (six landings lost ten minutes each in the hour before, main through the coordinator). Self-tests: the lock is taken with its holder line and released, a stale holder is reaped and the lock retaken, a fresh holder is waited for up to the cap, and the branch-side merge carries master's rows and cells plus the branch's.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:43:37 +00:00
igneum-labs
d79e9daa5d Merge fin-v4-rows e0bcf3e9 into master (gate: green on e0bcf3e9, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 19:43:19 +00:00
igneum-labs
826c4335c6 Merge master into key-succession-pin-tool (the pin tool's half only: elf/prior/ as a copy of the served pair, the manifest without a prior block; the real pin 4c6d6d644 stays on key-succession-pin for tomorrow's H)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:40:50 +00:00
igneum-labs
4e17b34de6 Merge remote-tracking branch 'build/master' into lab-signing-20 2026-10-08 19:40:19 +00:00
igneum-labs
ba43b1cc97 P01 part A: hl-v6-all (the class v6 D1 object) a million vectors with zero disagreement on the 5090, 4090 and 3090
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:39:14 +00:00
igneum-labs
cfe22118d6 Merge box/master bd8223cfb into counter-asic-4-docs-3 (rule 26: master's map, harness page and registry taken, the lane's cell harness:economics-model re-added and the batch eco-d4-20261008-01 re-recorded; the export list both ways) 2026-10-08 19:39:06 +00:00
igneum-labs
1ca1eed751 Merge build/master into reference-apps: master's steward files carried, the reference-apps cells and batches re-recorded, pages rebuilt
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:38:55 +00:00
igneum-labs
7d3e1e321b P01 part A, run p01-20261008-01: hl-v5-win a million vectors with zero disagreement on the 5090, 4090 and 3090 (POW-01 in progress; the v6 packs, OpenCL and Metal remain)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:36:19 +00:00
igneum-labs
4edc8cf6aa Research lane D's family-gate cell and batch (POW-02, ROT-07, ROT-08 partial; recorded by the steward at the lane's ask)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:35:58 +00:00
igneum-labs
e0bcf3e9f5 Registry batch: FIN-08 through the finality-sim cell (RUNNING, the real-node rows joined to the simulator's; the combined-boundary runs are tomorrow's plan), so sim/results_v2.md and its rows move together
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:35:40 +00:00
igneum-labs
e7668becda The merge normalises master's registry copy on every landing; --check refuses a row outside the four decisions
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:35:04 +00:00
igneum-labs
062931f6e5 Merge master 259f799e 2026-10-08 19:34:12 +00:00
igneum-labs
fb3230c6e7 The master's status discipline in the writer: four decisions, the p.139 record, the four approvals, several records per case; the REV ids from the master; INT-07's V6-12; the two kills as FAIL
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:34:09 +00:00
igneum-labs
259f799ec3 Merge ci-rev-suite 1dd0576c into master (gate: green on 1dd0576c, recorded by tools/ci/pre-push.sh; landed on the box mirror under the exception declared by main: main's ruling, 7 Oct 2026 19:5x UK: the GitHub account is suspended, lanes land on the box mirror's master, the box gate stamp is the verdict; GitHub gets the fast-forward when it answers) 2026-10-08 19:32:51 +00:00
igneum-labs
2c5db21153 Merge remote-tracking branch 'build/master' into reference-apps-f04 2026-10-08 19:32:33 +00:00
igneum-labs
52c785e764 Class v6: the genesis dataset size DECIDED at 4 GiB (the founder's word at 20:00 UK, FROZEN as class_v6_dataset_steps [(0, 4096 MiB)]); change (3) and the served policy sentence as decided; the chip line's a-node-ahead end 3.3x with the floor marked
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Documents-only copy of counter-asic-4 248837578 over master f1c1fd9d1 (the 21:00 landing): the design and analysis documents, spec 05 and 06 (06 and the export list three-way merged), sim/economy/coexist (python, JSON, TSV), the map cell harness:economics-model added to master's map, the batch eco-d4-20261008-01 recorded through tools/ci/test-record.mjs on master's registry copy; nothing under igneum-pow, proto-cuda or the apps; the branch itself unmerged.
2026-10-08 19:31:58 +00:00
igneum-labs
6d28e93693 The master's wording: the three boundaries with 'automatically' (every page and the pins), the ranking ceiling in the master's words (no present-tense rank; what an independently substantiated pass would and would not support), the proving record served as a disrupted record, not a benchmark
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:30:45 +00:00
igneum-labs
75e91cc4d4 Registry: the node lane's rule v4 real-node run (fin-v4-realnode-20261008T2025Z on 12424341) onto FIN-02 and FIN-07 through test-record.mjs; the evidence under sim/finality-attacks-results/
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:28:51 +00:00
igneum-labs
af05213034 test-record: a batch cell may narrow its write to named cases ("cases": [...]) so one case's incident never writes a cell's other cases
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:28:07 +00:00
igneum-labs
d77336eedb The deploy script records its own pid before it starts and refuses a second run (the one Mac-side publish path); the chip per-tier line in the three-row form with the placed 32-lane floor and P04 FAIL at both ends; the decided dataset clause verbatim
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:28:03 +00:00
igneum-labs
9f17873911 Merge outcome-ledger 2e6914f8 into master (gate: green on 2e6914f8, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 19:27:43 +00:00
igneum-labs
59d3a7283a The founder's word by construction: the kit ISA check, kill-by-name rule 5 and the no-kill shim, the cut batch's read-back
tools/ci/kit-isa-check.sh disassembles every executable of a kit (objdump) and is red on any AVX-512 encoding (a zmm register, an EVEX opmask, an EVEX-only mnemonic); self-tested with a fake objdump; in the gate as a self-test, in merge-to-master.sh over any executable a landing adds under packaging/kits or bin, and for the shipper's cut gate over the kit. kill-by-name-check.sh rule 5: a pgrep -f or pkill -f pattern passes only when anchored on the exact command line (^/full/path or ^command), the bracket form, a variable, -x or -F; a bare path, a log name or a word is red; the tree reads clean under it. tools/ci/no-kill-shim/{pkill,killall} exit 97 with the pid-file rule when first in PATH (the one line for a shell rc is in the file; the founder's own shell is his to change). test-record.mjs refuses a batch that declares a cut without the binary's build-N:/srv path, its commit string read back equal to the manifest sha, and the kit check's clean line.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:27:43 +00:00
igneum-labs
8cddf8e273 Merge master into fin-v4-rows
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:25:55 +00:00
igneum-labs
955e1256c7 Merge branch 'ci-rev-suite' into ci-int-suite 2026-10-08 19:25:53 +00:00
igneum-labs
1dd0576c2c merge-to-master: the harness page regenerates after the registry rebuild, never from a copy with conflict markers (the REV landing's 20:24 loss); fixture extended
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-08 19:25:37 +00:00