Three classes from the 0.3.17 night. (1) bash reads a script incrementally: tools/build-remote.sh was edited while a four-minute remote
build ran, the running copy continued at shifted bytes and died with a syntax error after the build had succeeded on the box; the
four long-running tools (build-remote, cross-remote, workers-remote, move-hand) now keep their body in one brace block ending in exit,
parsed whole before a line runs; tools/ci/whole-body-check.sh (in the gate, self-test with a block-less copy) holds the shape.
(2) A fork build pairs with the igneum-pow of the igneum worktree it sits in: a fork at 12153428 under a master worktree failed in
kaspa-pow four minutes in (no chain_program_shadow; master's igneum-pow predates release-0.3.17's); build-remote.sh says the
pairing on its first line ('pairs with igneum 6f8d7a7e (detached): igneum-pow 0.2.0') and the JSONL line carries pairs_with.
The first version of that line used '[ -n ... ] && echo' inside an assignment's $( ) and set -e ended the script on the false
status; fixed. (3) move-hand.sh restart <hand> [--digest <hex>] [--go]: after binary installed a release, restart ONE unit and read
it back (first exec line, commit string in the running binary, digest against the wanted one, igneum_getNodeInfo powEngine over the
node's loopback EVM RPC); the digest readers tolerate a missing line.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Main's order of 7 October 2026 after RunPod's Ubuntu 22.04 canaries (glibc 2.35) refused the box's GLIBC_2.38 binaries and HiveOS
turned out Ubuntu 20.04 based (2.31). The table lives once, in tools/ci/glibc-ceiling-check.sh (--class, --ceiling-of; self-test
covers the table, an unknown class and a 2.34 need against hive); tools/build-remote.sh --ship hive|rig|seed|linux (default seed)
and tools/workers-remote.sh --class (default rig) build with zig at the class's glibc and check against it. provision.sh installs
docker.io (user build in the docker group) for the proof. Proof: fork 3bfe346f at class hive, igneumd and igneum-miner need
GLIBC_2.30; the workers at class rig need GLIBC_2.17; all four run in ubuntu:20.04 (ldd 2.31) and ubuntu:22.04 (ldd 2.35) and
print their version or usage lines (the OpenCL worker its own no-libOpenCL message, the binary running in a GPU-less container).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The previous commit's 'the workers at 2.36' was not yet true: the first zig build died on __isoc23_strtol because -I /usr/include
for CL/cl.h put Ubuntu's glibc 2.39 stdlib.h in front of zig's bundled 2.36 headers. Fixed; proof on the box (6 s):
igneum-worker-cuda 6,759,272 B sha256 690c8e91..., igneum-worker-opencl 307,264 B sha256 329fb6a9..., each needing GLIBC_2.34
and libc only, the ceiling check passing. The two static flags zig ignores are dropped on the zig path (zig links its libc++ in).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Main's order of 7 October 2026 after a seed took 14 restarts and three minutes down on a glibc 2.39 binary. provision.sh:
step_zig (zig 0.17.0 from ziglang.org, sha256 from the official download index) and cargo-zigbuild 0.23.4 in the cargo tools.
tools/build-remote.sh --ship [--glibc 2.36]: cargo zigbuild --target x86_64-unknown-linux-gnu.2.36 with zig as the C/C++
toolchain (the Mac's infra/cross/build-linux.sh recipe), artefacts from the target-triple dir, each checked by
tools/ci/glibc-ceiling-check.sh (need at most the ceiling; self-test fires on 2.38 against 2.36, passes 2.34 and 2.36;
--symbols reads a saved objdump -T text so CI needs no ELF tools). tools/workers-remote.sh builds the two GPU workers with
zig at 2.36 by default (GLIBC=native for clang). Proof on the box: fork 3bfe346f igneumd needs GLIBC_2.34 (47,023,120 B,
sha256 345dfb95...), igneum-miner GLIBC_2.34 (9,248,808 B, d09dc27b...), 3 min 17 s cold through zig; the workers at 2.36.
Rule: anything that ships to a seed or a HiveOS rig is built with --ship; a plain build (glibc 2.39) is for the box and
Ubuntu 24.04 hosts only.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The X9 made the k = 0.33 column of the chip model a product class, so
every public sentence that stated 2.1x alone now states 2.1x (k = 1) to
3.9x (k about 0.33), with the ladder's second rung taking the X9 bracket
to about 2.8x. The ladder lane's litepaper paragraph 'The work that waits
can grow' and ledger row M34 are taken from branch ladder (7003f9f, those
two hunks only) so the ladder is on the site before the code ships.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The reset itself has been master's behaviour since 3e6a488 (remote-run.sh checkout_tree: git checkout -- . and git clean of the
overlay files, target dirs and stamps kept, before the branch checkout); the UI lane met the class again from worktrees whose
tools predate it (remote-run.sh is piped to the box from each worktree per build). The check reads checkout_tree() and fails
when the reset and the clean do not both come before the branch checkout; self-test: the real script passes, a copy without the
reset fails, the same lines moved after the checkout fail. cargo-audit: already owned by provision.sh step_cargo_tools
(1eec354), confirmed ok (0.22.2) on the box; nothing added. Gate GREEN, 33 checks.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The fleet's 22:09 UK incident (a Mac-side pkill -f <log file name> matched nothing, the roll-everything script lived on and wiped a held box) and the day's two pgrep self-matches are one class. The check flags pgrep -f / pkill -f with a plain literal (every one on a line), any pgrep/pkill on a file-name shape, and ps | grep with a literal; it allows the bracket form, -x, -F pidfile, kill $(cat pidfile), a variable and a full path; 11 banned and 16 allowed shapes in its self-test; 0.15 s over the tree. The 25 pkill -f sp1-gpu-server inside bash -c bodies (which matched the calling bash) are pkill -x; the other 11 literals take the bracket form; prover-socket-check accepts both. Row R in the record; the CLAUDE.md rule names the check and covers pkill and file names.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Bitmain's Antminer X9 (1 MH/s at 2,472 W, about USD 5,600) ships from
July 2026 and RandomX 2.0 shipped on 25 March 2026. Five sentences on
the home page and the litepaper that said or implied RandomX is
chip-free now state the X9 and its date; sentences that only name the
technique stand. Ledger row X34, C2 closed by the fact, the text check
carries the new sentences.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The box's 34 red rows of 6 October classified (docs/analysis/ci-failures-2026-10-06.md section 6): 22 iterations, 12 in three real classes (instant deaths with nothing kept, a shared worktree directory, an unread dry run). remote-run.sh now: pre-flight (subcommand, manifest, -p package, --features) refuses in a second with exit 3 and a class; the last 400 lines of every run kept in /srv/builds/_log/runs; a class on every row (compile-error, link-error, test-failure, instant, no-test-matched, slot-timeout, no-dir, preflight-*); a cargo test whose filter matched no test exits 3; a per-worktree lock in checkout and run mode; every red row appended to /srv/ci-red/red.jsonl as source box. red-watch.mjs never posts a box row alone and sends one digest a day (counts per class with each class's guard); the timer runs tick. Shared group cired on the box so the runner and build append to one file. Shown in a sandbox on the box: pass, failing test, empty filter, bad package, bad feature, missing subcommand, compile error, broken manifest, two concurrent runs of one worktree.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The owner, 6 October 2026: "Im not sure about the site can we revert it but polish it? minimal, simple but everything needed
for the dopamine and emotional triggers of the gpu miner", then "cant we have a dag animation like we had before? with the
shards making up the block, then the final line and the other bits that looked really cool but brought upto date?"
Restored (a new commit, no history rewrite): site/index.html as it stood at 575abef, the step-view page with its hero, facts,
scene, three things to check, downloads, build, wallet, journey, economics and the ledger band; nothing from tonight's other
work is undone (the litepaper, /live, the roadmap and benchmark wording, ledger X31 to X33, the scroll-zoom fix). Where the
page named a month it now carries tonight's sentences: the proofs card reads "Live rows arrive with the public testnet. The
public testnet is weeks away: three seed nodes and the public RPC are up, and it opens when the go checklist closes."; the
journey lead reads "Six phases, four public gates ... No calendar dates: each phase closes at its gate." and its inlined
phases are the gate-worded ones. No "August 2027", no month anywhere on the page.
The scene (site/live-steps.js): the original bits, every one driven by the live feed. A real block arrives from the right
with a glow and its chain number under it; its outline is grey while pending and ember once included; as its shards are
verified or paid, particles fly in from the edges and its quarter cells fill molten; when every shard is proven the block
turns ember and the caption says how many provers were paid; a locked checkpoint takes a ring and a ripple and the dashed
"final" line sweeps in from the right to it, drawn only while finality is active (the legend no longer says "final to its
left"; the wallet card's state word "final, checkpoint 5" is the wallet's own vocabulary and stays). A proof or a lock that
lands on a block already off screen re-enters from the right as its own event, so every step is seen when it happens. The
counters gain blocks per second. The caption follows the newest block that changed step.
The miner's triggers, each one element with live or measured numbers, none invented: a network strip under the facts
(hash rate, blocks in the last ten minutes, the latest block's age ticking every second with its chain number and word,
shards proven and paid with the last ten minutes), each value flashing molten when it changes; a card picker from the bench
table with the time a card alone takes to find a block at the live hash rate (RTX 5090 127.7 MH/s at 227 W and RTX 4070
28.8 MH/s at 76 W from the 6 October Ember Tune on the three-card Windows rig, RX 9070 XT 17.8 MH/s from the 5 October eGPU
entry, Apple M5 Max 26.7 MH/s from the first live swap; the RTX 4090 and RX 7900 XTX shown as not yet measured with the
bench table linked) with the pool note; the fairness line (graphics cards only, a new program every hour, your card proves
the blocks, 80% to the miner and 20% to the provers, nothing to anyone else); the download buttons with sizes; a Discord join
line (the standing invite, recorded in docs/community/discord-hooks.md so nobody asks again); the testnet sentence.
docs/fud-ledger.md: G4, C2 and X8 each gain a status line saying the restored home page carries their sentence again; the
ledger-text check guards them on the home page again (55 sentences, 0 missing).
Measured headless over 40 s against the live feed: 19 captioned transitions, the strip reading 2.3 GH/s, 562 blocks in ten
minutes, 8 s ago, 2,970 shards paid, the picker's times 18 s (5090), 79 s (4070), 2 min (9070 XT), 86 s (M5 Max) at that
hash rate, 0.92 blocks/s; no console errors at 1440 or 390 in either theme; no horizontal overflow. Captures in
docs/plans/site-ui-3-shots/after-v2/: home-{1440,390}-{dark,light}-fold.jpg, home-{1440,390}-{dark,light}.jpg and
home-steps-1440-dark.webm (ten seconds of the scene). Checks: identity grep 0 hits on served files, link check 934 links
across 16 pages 0 broken, ledger text 55 of 55, contrast 32 pairs 0 under 4.5:1, api tests 5 pass. Not deployed: the owner
sees it first.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
