igneum-network/igneum
1
0
Fork 0
Code Issues Pull requests Releases Activity
2390 commits 395 branches 1 tag 2.7 GiB
Commit graph

568 commits

Author SHA1 Message Date
igneum-labs
ffc3a54bdc Class v6 research lane B: the hardware future, five years out (first cut and the full report in one file) 
docs/analysis/class-v6/hardware-future.md: one table over PIM and PNM (HBM-PIM, AiM, LPDDR5X-PIM, UPMEM), HBM3E, HBM4 and the custom HBM4E base die, fine-grained DRAM on logic, LPDDR6, 3D DRAM, RLDRAM 3 and Folded Banks, FPGA with HBM2e, wafer-scale, CXL, optical I/O, chiplets and UCIe, and SRAM at N2: energy per dependent read, latency, cost per GB, availability to a non-hyperscaler, the k_read and shadow k bands, the edge per joule at zero shadow against the 5090 and the M5 Max, and the one layer of the four that blunts each with a number. The three findings: a 2 GiB SRAM full store on one N2 reticle is the strongest five-year chip (13x to 17x per joule, about USD 0.3 per MH/s) and no layer reaches it (layer 2 moves its capex, not its joules; the shadow at k = 0.5 holds 4.8x); per-bank PIM is blind (1.6 percent of reads in-bank at 2 GiB) and the custom HBM4E base die is the f = 1 chip with its controller in the stack (6.5x to 14x), untouched by layers 1 to 4; the denominator is the wrong card (the M5 Max at 0.78 microjoules halves to thirds every chip edge). Section 5 carries the k bands for the research lane's chip rows; section 7 the four decisions with defaults and deadlines; section 9 the sources with URLs and dates. tools/ci/export-exclude.txt: docs/analysis/class-v6 is research, not public export.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 10:17:50 +00:00
igneum-labs
3ff6de36a4 Merge live-shards-8 1c1f90e8 into master (gate: green on 1c1f90e8, recorded by tools/ci/pre-push.sh; landed on the box mirror) 2026-10-08 09:19:53 +00:00
igneum-labs
6cfed4c227 Merge build-server 4d55daf5 into master on the box mirror: the site deploy checks (served links, the regrouped /miners count, the pipefail fixes) 2026-10-08 09:18:01 +00:00
igneum-labs
1c1f90e86d Live proving feed: the page says when its shard states are stale, and a chain block with an empty plan reads "no shards in this block" (8 October 2026, main's order: the project lead wants to watch proving and shards in real time; a stale snapshot must read as stale). Reading of the day: /api/live's proven and paid states come from the Devnet 3 observer's own node on igneum-build-1 (igneum-observer-node-dn3, EVM 26850), not from hub-1; that node stopped crediting payouts at 23:57Z on 7 October (paidShards frozen at 7,304 while blocks carried 8,457 proof records by 03:00Z and hub-1 counted 11,861 by 08:56Z), and the page kept showing the paid total as fresh. site/lib/proving-health.mjs (tests known-failed first: "Cannot find module", then 10 green on build-2): stale when the observer's last status read is over 60 s old (status_at, new in the observer's live_state.proving), when the node's tip trails the newest block by over 300 DAA, or when blocks carried proof records in the last 10 minutes and no shard moved to verified or paid for 10 minutes (the blind-to-records class); a quiet chain is not stale for lack of credits. /api/live carries stale, stale_reason, status_age_s, tip_lag_daa, records_10m and last_change_at on proving (two more indexed readings; scene/feed-contract.json lists the keys; the recorded fixture carries them). /live: the Proven tile's line reads "proving feed stale: ..." in ember, else "... last credit N s ago"; the inspector's count adds "feed stale". Scene 2.0.7: a chain block whose plan is known (block.number set, carried as plan on the normalised block) and empty reads "no shards in this block"; loading and late stay for an unread plan (shard-words test, known-failed on 2.0.6). Gate: the health test joins the site unit-test line. 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 09:12:56 +00:00
igneum-labs
6c865ce08d Merge spec-scrub 1d329ebd into master (gate: green on 1d329ebd, 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 09:04:07 +00:00
igneum-labs
1db6b43868 Merge gh-igneum-dir dd954178 into master (gate: green on dd954178, 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 08:57:37 +00:00
igneum-labs
1d329ebd70 fresh-repo.sh: IGNEUM_NEUTRAL_IDENTITIES keeps a neutral identity as it is (the public export's "Igneum contributors"), excepted from the mailmap and the verdict 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:53:52 +00:00
igneum-labs
032edfa2ce fresh-repo.sh: the founder's names come from the private list only; a history identity is mailmapped, never mined for names (the public export's "Igneum contributors" would have become a first name to rewrite) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:52:39 +00:00
igneum-labs
dd954178b8 Every gate gh call reads Igneum's own gh directory, never the founder's (main's rule, 8 October 2026, 10:0x UK): tools/ci/gh-env.sh sets GH_CONFIG_DIR=~/.config/gh-igneum for the gate, the hook, the merge tool and the CI-state reader 
At 09:46 UK the founder's gh had his other account active (his own work) and tools/ci/gh-account-check.sh refused every Igneum push from the Mac (the v5 lane's 56a50160 held local). The check now reads Igneum's directory: empty, it refuses naming the one step (the founder or main stores the Igneum token there with `GH_CONFIG_DIR=~/.config/gh-igneum gh auth login --with-token`; no lane does); another login, refused and named; the stored entry, passes; while tools/ci/github-suspended stands it skips with a line, because no gh call can succeed and the hook already refuses GitHub pushes, so the lanes land on the mirror meanwhile. Known-failed first: an empty directory, another login, the founder's directory never read (the fake gh records the directory it was given), the marker's skip, no gh. IGNEUM_GH_CONFIG_DIR overrides the path for the self-tests.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:51:55 +00:00
igneum-labs
4d55daf536 Site deploy checks: grep -c in place of grep -q on every piped read (grep -q exits on the first match and the writer fails under pipefail) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:44:32 +00:00
igneum-labs
c7bb10dee9 OpenCL worker: on an Intel platform rotr_var is rewritten to the shift form before the build (the Intel rotate fold, the Arc B580 bisect of 7 October 2026) 
Intel's compiler turns rotate(x, (0u - n) & 31u) into a rotate LEFT by n: lane 0's register trace on the B580 diverged
at instruction 6 of iteration 0 (rotr) and nowhere before, in both exchange modes, with every other family and the
dataset kernels bit-exact. proto-opencl/intel_rotr.h rewrites the one helper line when the device's vendor or
platform string holds Intel (host.c's buildProgram and the prepare path), no other vendor sees a change, no pack or
consensus text moves. proto-opencl/test_intel_rotr.c (the pre-push gate runs it) feeds the line through the rewrite
under Intel, AMD and NVIDIA strings and asserts the outputs.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
(cherry picked from commit c6aa376a5d)
 2026-10-08 08:44:23 +00:00
igneum-labs
917af9bf22 Site deploy checks: the /miners row count reads table#bench-current (the regrouped bench), known-failed first (an empty current table reads 0), floor 20 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:44:00 +00:00
igneum-labs
686b757a97 Site deploy checks: pages read into a variable before grep (curl | grep -q under pipefail read false on a match: three false DEPLOY REDs) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 08:36:59 +00:00
igneum-labs
4b603b28f5 Merge ca3-v4-amend 2fa54cc6 into master (gate: green on 2fa54cc6, recorded by tools/ci/pre-push.sh; landed on the build mirror) 2026-10-08 07:11:38 +00:00
igneum-labs
1d48fc3f98 CA4 hot-table job: the first kit flattened the eight hot packs into one folder (the job exited 2 in 0 s, pack hot32k4 missing); kit b carries packs/<name>/ for all nine, the fetch and run ids take a b suffix 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 06:30:27 +00:00
igneum-labs
a5a0a687c4 Merge ca3-v4-amend 11f88bdd into master (gate: green on 11f88bdd, recorded by tools/ci/pre-push.sh; landed on the build mirror) 2026-10-08 06:29:02 +00:00
igneum-labs
11f88bdde6 PC 1 Ember per-card job: a card the app cannot steer stores one baseline row and logs no progress line; count the stored curve before calling the run empty (the 9070 XT run exited 1 on a real row) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 06:18:25 +00:00
igneum-labs
9d85a9f29c Merge ca3-v4-amend fa420559 into master (gate: green on fa420559, recorded by tools/ci/pre-push.sh; landed on the build mirror) 2026-10-08 06:16:47 +00:00
igneum-labs
fa420559b3 PC 1 Ember per-card job: the curve line cast an integer plus a string and threw (the 06:0x 5080 run's rows lost); string interpolation instead, and every card's stored curve printed at the start 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 06:06:24 +00:00
igneum-labs
03a62fc514 CA4 microbench job: the kit's class v4 pack as the --pack argument (the exe's parser demands one before the microbench branch; the probes read none); republished as -c after the sparse run 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 04:37:08 +00:00
igneum-labs
cf09054040 CA4 packs and hot-table jobs: $pid is PowerShell's read-only automatic variable (the 04:04Z packs run threw at its first pack); $progId instead; the packs job republished as -b 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 04:11:17 +00:00
igneum-labs
d6a278e465 CA4 microbench job: the installed worker's NVRTC DLLs copied beside the fetched exe and the exe run from its own folder (the loader refusal the research lane named) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 04:06:36 +00:00
igneum-labs
26d722fc04 CA4 microbench job: the research lane's exact command (no --device flag), every line echoed when the exe refuses (the 03:54Z run exited 2 after one unseen line); republished as -b after the packs 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 04:04:53 +00:00
igneum-labs
27b0adcae9 CA4 SM-sparse rerun (c): the research lane's ca4sparse5 exe (the parameter-capture fix); the card-free check's NVRTC compile result gates the run 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 03:56:27 +00:00
igneum-labs
8ebf714d9e CA4 SM-sparse rerun (b): the research lane's ca4sparse4 exe (the race-order fix); the job runs its card-free --list-race check first and stops with nothing run when the sparse variant is absent from the race order 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 03:05:51 +00:00
igneum-labs
8df8943f7a Merge github-suspended 7942ac70 into master (gate: green on 7942ac70, 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 03:04:36 +00:00
igneum-labs
c52c867811 Merge ca3-v4-amend 746099e3 into master (gate: green on 746099e3, recorded by tools/ci/pre-push.sh; landed on the build mirror) 2026-10-08 03:02:03 +00:00
igneum-labs
7942ac70f4 merge-to-master self-test: the nested refusal call's exit is caught (under set -e a plain assignment of a failing substitution ended the script before any check ran) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:59:43 +00:00
igneum-labs
4a933e5b6a merge-to-master: the suspension refusal lets --self-test through to exercise the refusal itself (the first cut refused its own self-test under the real marker) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:58:49 +00:00
igneum-labs
746099e3f2 PC 1 measurement pass: the step budget keeps four minutes of the cap for the restore and sizes each step from the last rate read; a probe dev line answered in helper.log counts as the helper up when its heartbeat stays stale; the gate check pc1-step-budget-check.sh with the known-failed case (the third 5090 pass of 8 October, ended by its cap at the 300 MHz step with the card left locked) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:56:18 +00:00
igneum-labs
7ee6ac2855 PC 1 efficiency pass: the batch count follows the last rate read, and a step that would end within four minutes of the cap is not started (the third 5090 pass's cap killed the script before its restore) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:54:13 +00:00
igneum-labs
83d4b6cacd No push, fetch or poll of GitHub while tools/ci/github-suspended stands (8 October 2026, 03:5x UK): merge-to-master.sh refuses a GitHub remote with the line and exit 2 before any gh or git call; the pre-push hook refuses any push to a GitHub remote 
At 03:39 UK a lane's default-remote call of the merge tool polled GitHub's CI 21 times (403 each) before it was killed by pid and re-run with --remote box; the rule since the suspension (never touch GitHub) rested on every lane remembering the switch. The marker is tracked and dated; main removes it at the cut-over. Known-failed first in both self-tests: a GitHub remote under the marker is refused with the line naming --remote box (exit 2 in the tool), a mirror remote is not, and no marker means no refusal.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:52:31 +00:00
igneum-labs
63f7ae204b PC 1 helper scripts: a probe dev line answered in helper.log proves the helper up when its heartbeat file stays stale (the 02:45Z restore answered with helper.alive stale past 60 s) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:47:46 +00:00
igneum-labs
71838a01f3 PC 1 clocks restore: the cleared helper sequence (empty cmd.txt, task restart, heartbeat) before the rgc; the third 5090 pass hit its 45 min cap at the 300 MHz step 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:43:08 +00:00
igneum-labs
2423ecf986 Merge build/master (360eb152) into crypto-engage: funding.md conflict resolved to the in-house pass with the pre-public scrub applied; in-house-pass.md scrubbed the same way 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 02:22:08 +00:00
igneum-labs
8329f91e40 Merge remote-tracking branch 'build/master' into ca3-v4-amend 2026-10-08 01:58:08 +00:00
igneum-labs
6d95617a0a CA4 SM-sparse rerun: the research lane's exe with the race fix (ca4sparse3, sha256 0ba97edc...), the rows carry served variant and sparse_blocks and mark a base-served sparse row FAILED; new fetch and run ids 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 00:47:32 +00:00
igneum-labs
7e03da7719 PC 1 queue: the hot-table streaming-hint job (the eight packs-ca2-hot packs plus the mx8 control, base against --variant ldcs, unlocked and at the knee through the cleared helper sequence, rows labelled ca4-research) with its kit and publish lines 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 00:25:06 +00:00
igneum-labs
b9626407d9 PC 1 helper scripts: the 0.3.20 helper is started over an empty cmd.txt and its heartbeat waited for (the update-return lane's cleared sequence, 01:17 BST); repeated when helper.alive is stale; a refused start (0x800710E0) or no heartbeat in 60 s stops the lock path and reports; the padding removed; the 5080 grid republishes as -d 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 00:23:23 +00:00
igneum-labs
b94da5956f pc1-publish: the CA4 packs kit title restored (a regex edit had cut it at the quote and broke the script) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-08 00:12:05 +00:00
igneum-labs
2cc8f8b31c identity-check: PC 2 is rewritten card-free as "the second Windows rig" (its cards are in dispute between lanes tonight; PC 1's verified list stays) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:58:53 +00:00
igneum-labs
58f6b09751 CA4 packs job: the two per-load rows carry construction=unsound-energy-reading-only (the research lane's verdict, 22:44 UTC) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:46:29 +00:00
igneum-labs
815344fb82 Merge build/master 8a42340f into spec-accept-23 (the hash lane's derivation recipe and test taken; the rewritten 1.4.6 keeps its program-id paragraph, which carries the recipe's text; AP-F8-4 ordered before AP-F8-5; the public ledger and the ledger page regenerated) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:44:49 +00:00
igneum-labs
30713a67b0 Spec read-back and the ledger page (AP-F8-5, 7 October 2026 night) 
Spec 01 1.4.6: the constants table gains ACCEPT_TAG, LANES, the (c'') window-cap literal and the first-refused reading of MAX_SATURATED. tools/ci/spec-constants-check.mjs joins the gate: every "Constant | Value | Where" table of spec 01 against the crate's pub const items (an expression over constants evaluated, a byte string compared, a pending row skipped while the constant is absent), known-failed fixture first. igneum-pow/tests/spec_readback.rs derives every pinned id of the "Seed | Attempt | Id | Note" table through the crate and draws each class v4 row with its era to the stated attempt; a must-differ row asserts another id (box suite on the branch: 64 + 7 + 4 + 19 + 2 + 7 + 2 passed, build-remote rc 0).
tools/ledger-page.mjs: both entry heading forms render (M1 and AP-F8-1; the single-letter form dropped every in-house pass row from /ledger), the pass gets its own section, a known-failed self-test in the gate. docs/fud-ledger.md drops the USD 50,000 prize clauses and "paid independent cryptanalysis" (decision item 1 of 6 October: no bounty; the in-house rule of 7 October), so the regenerated site/ledger.html carries neither. checks.txt regenerated (73 checks).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:44:14 +00:00
igneum-labs
8e8f0b60cb CA4 packs job: the per-load shadow's fixed export (mx8_shl256x27_v2, id bd64b207a30413fb) runs beside the first (the known-failed record); every row names its pack directory and program id 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:31:11 +00:00
igneum-labs
3509319a2e PC queue: the class v5 Intel fingerprint job moves to PC 2's Arc B580 eGPU (main's word; on the shipper's PC 2 clear only), the v5 kit fetch targets both PCs 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:25:07 +00:00
igneum-labs
50db9c4bae PC 1 queue: the v5 lane's own Intel Arc B580 script (class-v5 a4b08245) in place of my Arc variant; its publish line 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:24:25 +00:00
igneum-labs
1f83f7738c PC 1 queue: the class v5 kit fingerprint jobs (the v5 lane's 9070 XT script as written; an Arc variant that takes an Intel device by name from --list and reports its absence), the kit fetch and publish lines 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-10-07 22:22:40 +00:00
igneum-labs
e55d2aaefe The founder's name out of the ledger row and two script comments (the pre-public scrub check of the gate); the mirror's master 7bd0420d merged in 2026-10-07 22:21:37 +00:00
igneum-labs
16839685e4 Merge build/master c862a233 into ca3-v4-amend (the ledger keeps both sides) and the AP-F8-4 ledger row (the program id derivation text; igneum-pow against a5d735b2 differs in the recipe refactor, the exporter's print and the new test only; program.json outside the digested pack bytes) 2026-10-07 22:11:07 +00:00