the project lead's decision of 6 October 2026: the Mac runs nothing the network depends on. docs/plans/hands-on-build-1.md plans node 1
and the observer (its node and tools/observer) as systemd units on the box, one hand at a time, with ports, DNS names,
exec recovery, rollback and the decisions for main; infra/build-server/hands/install-hands.sh writes users, dirs, run
scripts and units (inert, run on the box); move-hand.sh (dry run by default) builds 0.3.15 on the box, copies the override
and snapshot, rsyncs each data dir hot then stopped, starts the unit and prints the hand's first executing line, copies the
observer env by scp (mode 600), stops the Mac's observer first, unloads the launchd agents last. CLAUDE.md's running-agents
rule carries the decision. lib.sh: a per-worktree lock on the box across sync, build and fetch (the shipper's collision at
18:48:56Z). remote-run.sh: the clean-tree test excuses stamps and target dirs at any depth with --untracked-files=all (a
nested stamp failed every checkout of /srv/builds/igneum for twenty minutes) and removes a stale .git/index.lock; the
self-test covers both. Plan rows for the three and the cuda prover pair built for the shipper.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The fleet's 10-member load run (6 October 2026, 18:14Z to 18:24Z) accepted 0 shares: the pool fork's miner re-checked GPU
shares with a fixed class v2 program while the 0.3.14 workers hashed class v3 (docs/plans/pool.md section 9).
pool: node.rs builds the share verifier's seeds from the template's pow_epoch with the class and the era, installs the
node's genesis day, dataset size and class v3 activation beside the schedule, and sends program_class, next_program_class,
era_seed, era_index, genesis_day_index, genesis_dataset_log2 and program_class_v3_activation_daa in `seeds` and
program_class and era_seed in `job` (protocol.rs, additive fields); verify.rs tests through EpochSeeds::v2. Protocol,
vardiff, PPLNS, stats API and page otherwise unchanged. Suite 22 of 22 on igneum-build-1.
igneum-pow tests/recheck.rs: for class v3 (packs-ca2-mixer/mx8-devnet-epoch0) and class v4 (packs-ca3-v4/v4-devnet-epoch0)
the pack's program reproduces the pack's 96 vectors (the worker's reference), the chain seam the node engine calls
(Epoch::chain_program, chain_dataset_day) builds the same program, one known nonce hashes equal through both paths, and
the fixed-v2 program of the same seed disagrees; the pinned v3 and v4 program ids differ. 2 of 2 on igneum-build-1.
packaging/hive: the pool:// mode merged with master's per-card IDENTITIES=auto and OVERRIDE handling (selftest passes).
pool/README.md and pool.md: building on igneum-build-1 (the vendor/igneum-node symlink on the box).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
proving/igneum-prove depends on vendor/igneum-node-exec/igneum/evm-types, a member of the fork's workspace that inherits
from the fork's root manifest; syncing that one directory left cargo without a workspace root on the box. lib.sh now groups
path dependencies by git top level: a repository under vendor/ is pushed to its mirror (a fork worktree to
/srv/igneum-node.git, a repository of its own to /srv/<name>.git, created on first use), checked out whole at
/srv/builds/<worktree>/vendor/<name> and overlaid whole; run-from-mac.sh wires every vendor repository the Cargo.toml files
reach. libprotobuf-dev added (sp1-prover-types imports google/protobuf/empty.proto). build-remote.sh no longer fails on a
default artefact the caller's own -p selection did not build. Proof on the box: igneum-prove-host 71,943,192 B, sha256
e9213e3a6c979512d7859f6d8e848105bab53f4355e99fb0d30fb4a72c2d5714, 1 min 05 s warm. Plan: gotcha rows for the case, the
protoc miss and one lost ssh session (collector cleared by test).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Mine, Earnings and Settings on igneum.network/miner are now the installed 0.3.14 app on PC 1 (RTX 5090 at 122 MH/s,
222 W, 0.55 MH/W, 56 C; RTX 4070 at 28.7 MH/s, 76 W; RX 9070 XT at 18.9 MH/s, 198 W; 169 MH/s at 532 W; Ember Tune
and Power control on), shot read only by the signed run job site-capture-pc1-1 with Edge headless and brought back
over the relay. Masked in pixels: the job's own strip cut out of the content column, the rail foot and the Settings
name field (hostname, address) painted over, the Earnings address box painted over. Captions say PC 1, the time, that
no electricity price is set and that the tune lines are stopped tunes from before that afternoon's Power Helper fix.
Prove, light and the phone width stay the Mac's (PC 1's Prove shows an exporter error line that belongs in a bug
report; headless Edge followed the Windows theme so its light set is dark; its 390 px window clipped the right edge).
The contact sheet is refreshed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The node pin moves to 4c6b129d (release-0.3.14-node), which is what the live payload inputs carry, so windows-ci's payload-inputs step is green again (red since e9eca23 on master's 0.3.13 pin against the 0.3.14 inputs).
Conflicts: infra/fast-time/override-60x.json keeps master's side (the fresh-rule field was already there at u64::MAX; the release line would be a duplicate key); tools/ci/playbook-quit-check.sh keeps master's rule 2 and pre-rule list with the release side's Ember allow entry and the 0.3.15 expiry check.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
remote-run.sh gains a checkout mode used by lib.sh: git checkout -- . and git clean -fd (target dirs, the sha stamps and
ignored files kept), fetch, branch at the commit, then a clean-tree check; the overlay follows. Before this, the rsync of
uncommitted files stayed in the box's tree and the next commit's git checkout -B refused with 'local changes would be
overwritten'. remote-run.sh --self-test reproduces the dirty tree (edited tracked file, untracked file, target dir, sha
stamp), shows the plain checkout refusing and the mode landing clean on the new commit; it runs in ci.yml and passed on the
Mac and the box; a live dirty-then-clean pair on the fork worktree passed too. Plan: section 5, gotchas of the first day.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The "LIVE DEVNET" scene on the home page showed blocks 21, proven 0, locked 3 and grey nodes while the devnet stood past
chain block 140,000 with 2,090 shards paid and a checkpoint locking every 30 s. Four causes, all in site/index.html:
- "blocks" was the scene's own spawn counter, never the chain;
- "proven" counted blocks of the 90 s window whose shards were all paid, and proofs land a median 380 s behind the tip
(median_proof_lag_s), so the window can never hold one;
- "locked" counted locked checkpoint blocks inside the same 90 s window (always about 3);
- the live path skipped the colour state machine, so every real block stayed grey, and a failed first fetch left the
simulation running under a label that looked live, with no further polls.
Now: /api/live carries state.height (the chain block number, the same field /api/stats reports, one indexed subquery);
the scene's counters read chain block (state.height), shards proven (proving.paid_shards_total, the node's count of paid
shard records, with the 10-minute count and the median lag in the note) and last lock (finality.latest_locked_index);
live blocks take the observer's own words (pending, included, excluded, proven, locked checkpoint, final to its left),
refreshed on every poll, with a legend in those words; the hero tile shows the chain block instead of the node's
held-block count; polling never stops (2 s while fresh, 10 s while off) and the off state is labelled "simulated preview ·
live feed unavailable" or "observer offline, last update N ago" with a note that the counters count simulated blocks.
Also found while verifying: the scene advanced a fixed 16 ms per frame, so it ran at double speed on 120 Hz displays;
it now uses the real frame time, capped at 50 ms.
/live had the same window bug in its "proven 0/16" counter; it now reads the observer's 10-minute truth (chain blocks
with every shard paid, of the chain blocks planned in 10 min) with the lag and the chain total in its title.
Verified against the live API through tools/site-serve.mjs in the built-in browser and headless Chromium:
chain block 140,489, shards proven 2,130, last lock #6,784; the failed-fetch state shown by blocking /api/live.
Screenshots: docs/plans/site-ui-3-shots/item0/ (1440 and 390, live and failed). Checks: identity grep 0 hits,
link check 708 links 0 broken, site/api tests 14 pass, site build clean.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Real screenshots of the live 0.3.14 app on the Apple M5 Max (paused, Ember Tune off, as the project lead left it), shot headless
at 1440x900 and 390 px, dark and light, the rail foot, the address and the machine id masked: miner-dashboard.webp
(the hero, same name), miner-earnings, miner-prove, miner-settings, miner-dashboard-light, miner-phone.
miner.html: a page-by-page section (the card row cell by cell, three page shots, light and the phone), the lead and
the meta copy name the four pages and pounds a day at the user's price, the Ember Tune feat carries the app's one
sentence and the measured PC 1 rows from the 6 Oct log (5090: 311.0 to 226.8 W for 0.15% of rate; 4070: 106.0 to
75.6 W for none; the floor, not the optimum), the Power control feat (one approval, then the helper task, no prompts),
the dev fee moved to Earnings as in the app. index.html: the same hero shot, two pills. build.mjs: the caption no
longer says 'until 0.3.6' (0.3.14 still says Igneum Miner). The 9070 XT row, the market price for pounds earned and
the bench anchor (the run 6 entry is on the ship0314 branch) are marked as owed or coming.
relay/playbooks/site-capture-pc1.ps1: a read-only run job that shoots the installed app's UI on PC 1 with Edge
headless and drops the PNGs on the relay (no quit, pause, resume or settings change; passes the playbook checks).
docs/plans/site-miner-2026-10-06.png: before and after.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Main's decision of 6 October 2026. CLAUDE.md 'Running agents on this Mac': every Linux and Windows cargo build and every
Linux test suite goes to igneum-build-1 through tools/build-remote.sh and tools/cross-remote.sh; the PCs keep GPU and
Windows-runtime jobs; the Mac keeps macOS binaries, the DMG and Metal tests under the lock. -Wl,--no-insert-timestamp in
cross-remote.sh, proto-cuda/windows-node/cross-build.sh and jobbuild.rs (two box builds byte-identical, verified).
The empty-commit class: tools/ci/commit-string-check.sh (self-test in ci.yml, shown firing on a Mac worktree build and
passing on a box build) runs on every igneumd from the three build scripts; push-build-inputs.sh adds node.commit_full,
the PC job writes a minimal node/.git from it at extract and cleans kaspa-build-info on a new commit (4 jobbuild tests
pass, run on the box); cross-build.sh refuses a worktree and cleans on a new commit. Plan: second worktree's clean build
1 min 18 s with sccache 604 hits of 993.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
remote-run.sh runs on the box behind BR_* exports: slot, sccache stats, the RESULT line and one JSON line per run in
/srv/builds/_log/builds.jsonl (v 1, id, host, tool, worktree, crate, kind, command, target, branch, sha, label, agent, slot,
wait_s, queued_at, start, end, secs, exit, compiles, sccache, load_end, artefacts; written on success, failure and the 2 h
slot give-up), the label ending in '; agent=<name>' (IGNEUM_AGENT, default the worktree). The remote checkout is a branch and
build-remote.sh cleans kaspa-build-info (release profile) on a new commit, so the box's igneumd carries its commit hash
(no Mac worktree build does: .git is a file there). cross-remote.sh fetches the GCC 13 runtime DLLs beside an exe that
imports libstdc++-6.dll. The plan holds the three benchmarks: clean node build 1 min 27 s (Mac 12 to 18 min), incremental
7 s (Mac 2 to 15 min), Windows cross 1 min 44 s (Mac 4 min 49 s to 12 min 28 s), their consequences and the proposed rules.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
