On 47ede3f, by hand, the app half of the exec fix (fork exec-sync-0313 1f59c5d0 is the node half):
- app/igneum-app/src/prover.rs: igneum_exportSegments [n-1, n] for a shard and [first-1, last] for a segment,
never from 0 (on a restarted node the records below the restart carry zero roots and the exporter refused every
cut, the fleet 16:02Z); exec_boundary() reads igneum_getExecStatus.restartNumber (or the startedFrom text on a
0.3.13 node) and the prover claims no shard and no segment below it
- proving/igneum-prove/export: seeds the port from the export's preState (the node's account dump after the first
segment), checks its root against the node's there and replays from the next segment; without a dump the
restart-aware replay (execRestart) and the genesis replay stay
- tools/exec-sync/net.mjs (15 checks: the persisted state, the file, the wrong pin, another chain, the unreadable
flag file, the account-dump cut) and tools/exec-sync/reorg.mjs (18 checks: a 300-block reorg from the ring and
from the persisted generation)
- infra/fast-time/override-60x.json: the duplicate proving_v1 block from the 0.3.12 merge removed (the
consensus-core test fast_time_60x_file_is_the_devnet_at_60x failed on it)
The three UI files are untouched (byte-equal to 47ede3f); the igneum-prove-r0 tree is not in this cut.
Green on this tip (6 October 2026): cargo test in app/igneum-app 28 + 8; node --test app/igneum-app/ui 35; the 13
CI checks of ci.yml that run on this Mac; tools/exec-sync/net.mjs 15/15 in 97.2 s against this exporter and the
fork's igneumd (IGNEUM_EXEC_BIN, IGNEUM_EXPORTER).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The audit (docs/plans/miner-ui-3-audit.md, 27 screenshots): no money anywhere, three red n/a per Apple row,
tuning split across two pages, 370 px of controls per card, Prove's 95 words and four zeros, Node's eleven
numbers, the jobs table on every Updates page, fixes two taps away, no light mode, no layout under 900 px,
developer lines on user surfaces.
The design (docs/plans/miner-ui-3.md): four sections (Mine, Earnings, Prove, Settings); the card row carries the
state word with its reason, MH/s, W with MH/W, £ a day at the user's electricity price, °C, Tune, the switch,
and its details under a chevron (cap slider, identities, pin, telemetry, the tune table); the tune line per row
(not tuned yet / tuning: step k of n with a bar / tuned N ago · point · next check <weekday> / measured only and
why / pinned / stopped / fleet pause); Tune all; the goal (Efficiency, Balanced, Maximum) with its £ a day; Power
control as one switch where it is the fix; the node as one line with Details; updates as one card; earnings money
first, IGN second; proving as a tier sentence per card; every costly action confirmed in place (quit, change
address, show key, trust mode), no dialogs; light and dark; a bottom tab bar under 720 px.
The build: index.html, app.css, app.js rewritten on the same engine routes; the pure blocks keep their API and
carry ember-tune's tuneNotice, tuneWord, toggle states and tune-line test (e9738dd) so the merge is clean; the
update card is named Igneum Miner; the Rust side is untouched. node --test on the four UI files: 35 pass.
Screenshots of the result: docs/plans/miner-ui-3/n00 to n26.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The console-window class (the project lead, 5 October 2026: "Windows Command Processor" windows on PC 1 whenever a remote job runs).
Measured on PC 1 (ae432dc7, Windows 11 Pro 26200, default terminal "Let Windows decide" = Windows Terminal 1.24) with
tools/windows/console-watch.ps1 (job run-20261005-182528): no child a job script starts from the app's headless
console opens a window (powershell, cmd, query, curl, nvidia-smi, wsl --status, a distro, interop cmd and powershell,
powershell -WindowStyle Hidden: 0 windows each); Start-Process in a new console opens a Terminal window (the known-failed
case: 2 windows), the same with -WindowStyle Hidden opens none (the known-finished case). The elevated path
(Start-Process -Verb RunAs -WindowStyle Hidden through the AppInfo service) is the one road left; its watcher
(console-watch-elevated.ps1, job run-20261005-184610) was cancelled at the UAC prompt.
- platform.rs: elevated_ps_line + elevated_command build the one PowerShell line every elevated launch uses (the NVIDIA
power cap, the sweep helper, the clock sync, an elevated remote job), -WindowStyle Hidden by construction; unit
tests on the line, the quoting and the Command.
- jobrun.rs: the elevated job path uses it; the relaunch helper's Start-Process carries the reason it has no
-WindowStyle Hidden (igneum-app.exe is a windows-subsystem program).
- tools/ci/windows-spawn-check.mjs (+ ci.yml): fails when a Command::new in app/igneum-app/src is not quieted,
a creation_flags is not CREATE_NO_WINDOW alone, a Start-Process the Rust code writes lacks -WindowStyle Hidden or
-NoNewWindow, or host.cpp spawns without CREATE_NO_WINDOW / SW_HIDE; self-test on known-good and known-bad samples.
- tools/windows/console-watch.ps1, console-watch-bg.ps1, console-watch-elevated.ps1: the watchers (run jobs).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
PC 1, 6 October 2026 06:58:47Z: the app came up after the 0.3.11 publish, had the installer verified 44 s
later and sat on "installs at the next safe moment" with its slot at minute 35; the project lead pressed Install now at
07:00:48Z. The slot staggers a fleet through a NEW publish; a machine that was off through the publish has
nothing to stagger. Now: published_at + 3,600 s <= engine start => slot_ok, logged once. The other safe-moment
guards (node synced, miner idle, boundary, network drop) are unchanged. manifest::unix_from_rfc3339 with tests.
For 0.3.12.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
the project lead, 6 October 2026: on PC 1 the integrated card sat between the 5090 and the 9070 XT. The list now shows
usable cards first, ordered by the measured rate since the start in 5 MH/s buckets (no flicker), discrete,
external and Apple before integrated, then memory; removed and unusable cards last; ties keep the detection
order. The row signature follows the order, so a reorder re-renders. Three UI tests. For 0.3.12.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
One prover at 2.2% coverage never had 8 consecutive proven blocks (C47, 6 October 2026); claiming whole segments makes it complete about 1 segment in 75 instead of none. The choice is deterministic per key (FNV of first block and key hash) over the untouched whole segments inside their deadline by a margin (240 DAA or 1.5x the last segment's time); the per-block path stays as the fallback. Unit tests for the grid, the grouping, the margin, the attempted set and the per-key order; 120 app tests, 8 core and 9 host tests pass.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Source: the scratch engine's own log in collect ember-c35-collect-1 (06:59Z): 22:31:02Z '0.3.10 is available: downloading',
22:31:05Z 'update: starting the installer first ... ota-apply.ps1', and the installed app's 'quit:' at 22:31:06Z.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
serde_json's to_value refuses a u128 over u64::MAX (18.45 IGN) and state_json turned the error into json!({}). A paid shard is 1.23 IGN on average, so a proving machine's dashboard went blank about 15 paid shards after every app start. Unit test over the boundary; 114 app tests pass.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
cargo test --release -p igneum-app under the build lock: 113 + 27 + 8 passed, 0 failed (the prepare_packs_tests,
resume_tests and provedefault tests among them).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The known-failed case, from PC 2's 0.3.9 log (run win-1ccfe586-20261005-200114): 1791234223 pause -> 'stopping the
miners (paused)' (every slot's restart_at cleared, the 5090 'off'); 1791235511 '[ok] mining resumed'; then
'0.00 MH/s, waiting' at every 30-s status line until the 0.3.10 restart at 21:49:41Z. Cause: Cmd::Resume re-armed
only slots whose watchdog said faulted; the 5090's slot was healthy and stopped, so nothing restarted it. The test
the_pc2_resume_of_21_25_11z_restarts_under_the_new_rule_and_not_the_old encodes that slot (faulted false, live
false): the old rule returns [] (the defect), the new rule [0]. cargo test -p igneum-app resume: 3 passed;
provedefault: 6 passed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
the project lead, 5 October 2026, 22:45 BST: "make sure we have ember tuning every single card for efficiency out of the box, the
more data = the better the tune, make an awesome system." Built on lever 3 (docs/plans/miner-eff.md), lever 2's signed
tuning section (docs/design/miner-tuning.md), the AMD telemetry helper (423936b, its --tune/--set-gmax/--set-plimit/
--reset contract) and the Power control switch (057f0ec). Design, data flow, tiers and the privacy line:
docs/plans/ember-tune.md.
- src/ember.rs (new): two knobs per card (power limit %, core clock cap MHz; memory clock never touched), the full plan
(power ladder 100..50%, then the clock ladder 90..60% at the chosen power), the confirm plan (the fleet prior and one
neighbour), the baseline plan (measure only), the marks (faulted, hot, memory_clock_dropped, unapplied, no_readings),
the choice (best MH/W within 1% of the top rate, then rate, then draw), the fleet record (a hash of the install id,
no address), the prior lookup and the kill switch (tuning.ember), the state machine on a fake clock. 9 unit tests.
- engine.rs: tick_sweep schedules every NVIDIA, AMD and Apple card (120 s steady, 600 s to the boundary, no job hold,
no pause, weekly, again after a driver major or program-class change, never under the manifest kill switch); the
probe (nvidia-smi clocks.max.gr + driver_version and the direct/helper mode; igneum-gpu-telemetry --tune for AMD);
tune_apply (nvidia-smi -pl / -lgc 0,<MHz> / -rgc directly or through the helper; the AMD helper per request);
Cmd::TuneProbe, Cmd::TuneSet; faults from rejected and mismatched hashes mark the step; the TUNE lines and the TUNE
{json} record, uploaded with the log; the Tuned line on the card state. The NVIDIA helper starts only with Power
control on: the --sweep job never counts as permission (no prompt on a PC with nobody there).
- sweep.rs: the helper protocol gains lgc/rgc (clock cap and reset) and resets the clocks after 20 idle minutes.
- state.rs, config.rs: the tune fields (clock cap, driver, class, source, the Tuned line); the nvidia-smi telemetry
query carries clocks.gr and clocks.mem; the AMD sample line's plimit_pct and gmax_mhz are parsed.
- ui: "Tuned: X MH/s at Y W (Z MH/W)" with the point, the source and when; measure-only cards say why; the Ember Tune
switch; tune-line.test.mjs.
- relay/lib/ember.mjs + relay/test/ember.test.mjs: the aggregation per (card model | driver major | program class):
median point, MH/W, spread, samples, machines; five samples converge, an outlier does not move the median, baselines
make no prior, de-duplication, the manifest merge keeps lever 2's cards. api/console.mjs fn=tuning and
tools/console.mjs tuning; tools/tuning.mjs --priors [--write tuning.json] [--site] [--tuning-off].
- site: the fleet priors table on /miners (site/miner-priors.json), the lever text.
- relay/playbooks/ember-tune-pc1.ps1: the PC 1 run (second engine with --sweep from a scratch copy of the install).
Measured tonight: see the bench log entry that follows the PC 1 run. The 9070 XT left PC 1's bus at 20:40 UTC and the
5090 needs the administrator prompt the project lead cannot answer asleep, so tonight's PC 1 run is the baseline plan on the 5090
through the whole pipeline; the two-knob tune on both cards is owed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
the project lead, 5 October 2026: "if we don't have to ask then don't ask". The NVIDIA power cap and the efficiency sweep need
administrator rights (one UAC prompt); PC 1 raised that prompt for cmd.exe at every app start and every sweep attempt
(17:00, 17:30, 18:12, 19:04 UTC today, each cancelled unanswered after 2 minutes; the "Windows Command Processor"
the project lead saw).
- config.rs: `power_control` (default OFF on every machine); `sweep` default becomes off and is implied by it (an
install carrying sweep = true without power_control is migrated to off on load).
- engine.rs: `elevation_allowed(power_control, sweep_only)` gates the power cap (`power_cap_plan` builds nothing when
off, the card note says so), the sweep scheduler, Sweep now, the sweep helper; no prompt on quit (the limits reset at
the next reboot); no second prompt through PowerShell when the window host's prompt goes unanswered.
Cmd::PowerControl(on): on = ONE prompt at that moment (every NVIDIA cap in one step), off = nothing asks;
`power_control_after_prompt` turns a refused, cancelled or unanswered prompt into "power control off:
administrator rights were not given" (switch back off, sweep off, no retries). Unit tests: off builds no elevated
command; on + refusal gives the notice; rights given keeps it on.
- platform.rs: `elevated_failure` maps the launcher's exit 251 and the "canceled" wording to the prompt, any other
code to the step itself.
- server.rs: POST /api/power/control {on}. ui: the Power control switch with the line "Windows asks for administrator
rights once; the cap and the sweep need them", the note beside it, the sweep switch disabled while it is off.
- The clock-sync prompt stays behind the Sync clock button only (unchanged).
- tools/windows/power-prompts-off.ps1: the 0.3.9 job that switched PC 1's sweep off through the API it has
(run-20261005-192313: sweep True -> False; the 0.3.9 cap has no off switch, it asks at an app start only).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
the project lead watched the 9070 XT at 90% usage with its fans barely turning and the app could not say what it drew: the
draw, temperature and MH per watt line came from nvidia-smi only, and the earlier per-watt figure used the board
rating. proto-opencl/gpu-telemetry.c prints one line per AMD card per sample (bus from SetupAPI by the display
device's name, kind, name, watts, temp_c, fan_rpm, fan_pct, mclk_mhz, gclk_mhz, util_pct, source), built by
build-windows.sh against vendor/adlx (the SDK clone), shipped by make-payload.sh and push-inputs.sh. The engine
runs it with -l 5 beside nvidia-smi (Source::AmdTelemetry, tick_amd_telemetry), parse_amd_telemetry fills
power_w, temp_gpu, fan_pct, fan_rpm, mclk_mhz, util_pct and telemetry_at on the AMD card matched by kind and
ordinal, so eff_mhw and the dashboard's existing line show it; app.js shows fan and memory clock when present.
Tests: three on the parser with lines captured on PC 1 and the Mac fixture; the sysfs path ran on a fixture tree.
Measured over 20:27:45 to 20:29:41 UTC with both cards mining (docs/bench-log.md, under the 9070 XT ceiling table):
9070 XT 198.9 W (193 to 212), 64 C, 657 rpm, 2,505 MHz memory, 3,290 MHz shader, 100% busy, 17.73 MH/s =
0.089 MH/W; RTX 5090 307.6 W, 69 C, 44% fan, 122.30 MH/s = 0.398 MH/W.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Step 1: app/igneum-app/src/provedefault.rs decides once per install (NVIDIA 12 GB or more, WSL2 answering on
Windows, Linux native, Apple silicon off until measured), never switching an explicit on back off; the Settings
switch line and the tile line say why (5 unit tests). tools/proving-v1/pc2-prover-cost.ps1 is the PC 2 job
(5 min mining alone, 5 min with the prover, GPU memory and host RAM peaks, the sp1-gpu-server's SM targets).
Step 2: the host gains --mode chain (consecutive fixtures, each block aggregated with the previous block's
proof by recursion), --mode aggregate (the live aggregator over shard proof files, a run of blocks in one
process) and --mode verify-segment (the node's verifier against the pinned aggregator key); the app's prover
loop gains aggregate_once (spec 7.8). Eight consecutive live fixtures (blocks 81046 to 81053, node 1's export
at tip 81076) under proving/fixtures/chain/. tools/proving-v1/pc2-chain.ps1 is the PC 2 job (held).
Steps 3 and 4: tools/proving-v1/coverage.mjs (the proven-block share and the on-chain latency from one node's
RPC), tools/proving-v1/net.mjs (the fast-time 3-node harness on 29950+ with the known-finished and
known-failed cases of the chain rule and the unproven rule), the four proving_v1 fields in
infra/fast-time/override-60x.json. Spec 7.8, the 7.4 rows, the 5.3 sentence, docs/plans/proving-v1.md.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The Notices block takes the hot-plug notices as on gpu-hotplug (card added, not usable, removed). The View block
and the Mine rows, the first-run rows and the Settings cards show a removed card (dimmed, no switch, the row goes
after five minutes) and a faulty card (named in ember, the OS problem code, the reboot hint, no switch); the big
button and the counts take only present cards; the name tooltip carries the tool's code, the device, the platform
and the PCI address. view.test.mjs covers the two states. host.cpp keeps both the WM_GETMINMAXINFO and the
WM_DEVICECHANGE cases.
Build tooling: push-build-inputs.sh and build-job.mjs take --no-node (the app engine only, no node source, no node
build, no node tests), for an app-only PC compile.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
From 18:23Z both Windows workers (CUDA on PC 1 and PC 2, OpenCL on PC 1 after the 18:34Z node restart)
refused every pack for epoch 34 with "the epoch seed bytes do not give the pack's IGNEUM_SEEDW_INIT", and
the miner and the app restarted them every 5 to 60 s until 18:44Z and beyond. The packs were correct.
The generator retries a rejected candidate with seed || k_le32 (attempt_words); epoch 34's attempt 0 was
rejected (246 of 16384 final register values saturated, limit 163) and attempt 1 accepted, so the pack
carried attempt 1's words while pf_load (proto-cuda/nvrtc/packfile.h) derived the expected words from
the bare seed. Both workers also matched jobs to pairs by those bare-seed words, so even a loaded pack
of a retried program would have answered "epoch seed mismatch" on every job.
The rule, in one place per language:
- packfile.h: pf_program_words(bytes, attempt); pf_load reads IGNEUM_PROGRAM_ATTEMPT and checks the
attempt's words; the refusal says "program pack and its seeds disagree: IGNEUM_SEEDW_INIT is not
attempt N of the epoch seed ..." in plain words.
- worker.cpp and proto-opencl/host.c: a job belongs to a pair when the seed hex the node sent is the
pair's (pairIs); the compiled-in placeholder pack keeps the word comparison.
- igneum-pow/src/packcheck.rs: verify_pack_texts / verify_pack_dir, the same rule in Rust; the miner
checks every pack it writes with it before a worker sees it (vendor/igneum-node pack-loop branch).
Tests pin the attempt vectors of epoch 34 on both sides (one vector, two implementations), that
epoch 34 is attempt 1 and epoch 33 attempt 0, a known-good pack of a later attempt, a known-mismatched
(out of date) pack, and self-contradicting packs.
- proto-cuda/nvrtc/emu/packfile-test.c (+ .sh, in CI): pf_load on a known-good attempt-1 pack, the
checked-in attempt-0 pack, and the known-mismatched bare-words pack.
The app (app/igneum-app):
- watchdog.rs: PACK_OUT_OF_DATE_CODE 44, PackRebuilds (at most 3 pack exports per epoch, then the card
shows the reason), pack_refusal (the worker's "error 0 pack" line and the miner's "PACK OUT OF DATE"
line), pack_epoch_of; tests on the incident lines, known-good and known-mismatched.
- engine.rs: exit 44 exports the pack again before the restart instead of a blind restart, the strip
says "program pack out of date, rebuilding", the card and the log name the condition; at the cap the
card is marked failed with the reason and tries again in 10 minutes.
The relay (relay/lib/parse.mjs): PACK_MISMATCH; the card reads "pack mismatch, rebuilding (N refusals
in the tail, M restarts)" in `node tools/console.mjs machines` instead of a bare restart count; tests
on the PC 2 tail of 18:27Z and a healthy tail.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The one long page becomes a rail with six sections, a thin top bar and the status strip under it; the setup
screens and the key sheet stay. Mine: one big start/stop, the numbers, one row per GPU with its switch, hash
rate, temperature and power. Prove: the switch with plain words, the counts, the verifier, the pinned ids.
Rewards: the address with one Copy, the key backup card, another address. Node: the plain lines, the consensus
digest, the next switch. Updates: the version, the jobs. Settings: one switch or slider per setting with one line
of help. Every function that existed is placed, none removed.
Engine (three small additions, each with a unit test): node.consensus_digest from the node's own line,
node.consensus_switches from the override file, proving.program_id and aggregator_id from the host's --mode id.
Hosts: the window minimum is 900 x 600 on both. UI tests: view.test.mjs (10) joins notices and update-card in CI.
Proof: docs/plans/miner-ui-2.md and the 19 screenshots under docs/plans/miner-ui-2/, taken from a build of this
tree running on its own port against the devnet.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
PC 1 after Adrenalin 26.9.2 (5 October 2026): two AMD ICDs each listed the iGPU and the RX 9070 XT, so the 0.3.9
app showed five rows for three GPUs (gfx1036, gfx1201, gfx1036, gfx1201 plus the 5090), ran two workers on the one
9070 XT at 9 MH/s each, called the iGPU discrete, and re-enabled it when its key moved from amd:1:gfx1036 to
amd:0:gfx1036.
detect.rs: the Windows path is a pure assemble(Inputs) over nvidia-smi, the worker's --list and the adapter list.
dedupe_platforms keeps one entry per card per vendor: the fuller platform wins (most GPUs, then the newest driver,
then the first listed) and another platform's device survives only at a PCI address the winner lacks (without
addresses: the same code and ordinal); the dropped entries go to the log, never to a worker. resolve_name puts the
Windows adapter name on the row: by PCI bus (the adapter list now carries DEVPKEY_Device_BusNumber and Address),
else by the gfx code's PCI device ids, else by the card names in a gfx table (gfx1201, 1200, 1100, 1101, 1102,
1030, 1031, 1032, 1036, 1035, 1103, 1150, 90c, approximate), else the table's words, else the code. The code stays
in `code`, the key and the row's tooltip; the key is vendor:code with "#2" for a twin, no index. hotplug::pref_for
reads a 0.3.9 key (vendor:index:code) when exactly one matches; the diff matches by key and PCI address, then by
vendor and name or code. The iGPU is integrated and off by default; the choice survives a moved index.
proto-opencl/host.c: --list prints ", pci bb:dd.f" on the info line from CL_DEVICE_TOPOLOGY_AMD or the NVIDIA bus
and slot ids (needs a worker rebuild through proto-cuda/nvrtc/build-windows.sh and push-inputs; an old worker
still works, by code and ordinal).
Tests: detect::tests has the five-row PC 1 list as a fixture (two platforms, the three adapters with DEV_13C0,
DEV_2B85 and DEV_7550) and asserts three rows named NVIDIA GeForce RTX 5090, AMD Radeon(TM) Graphics (integrated,
off) and AMD Radeon RX 9070 XT (discrete, on, 8 identities), the keys, the console line, and the diff against the
one-platform list before the eGPU (iGPU moved, 9070 XT added, nothing removed); plus the list parser with and
without pci, dedupe with and without addresses, name resolution, twins' keys. cargo test -p igneum-app: 96 passed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
