wiki: R1's named instrument ran at 1080p on the GTX 1070, and the arm resolves there at the march's own cap (4b68606, ff8fb26, ea1daa4)
The front page still said the demo had no GPU upload and no frame timing and that R1 waited on hardware; the
hardware has now reported. r1_march_alone's camera path on buildbox's GTX 1070: cap 1 15.17 ms (inside 16.67),
cap 8 70.76, cap 64 91.15 - 5.5x over budget on the instrument the gate names, with the read-back near
cap-independent and the old 551.5 ms 46-pass row shown to be measuring the harness. R1_ARM_CAP then prices the
bound-skip arm at the cap the march actually uses: armed 70.519 ms against the +inf control's 73.945, paired
median 3.420 ms (+4.6%, 112.0 sigma, 14/15), and the 768-ray null survives a third instrument at -0.001 ms. So
R1 moves from unmeasured to measured-and-unmet, the arm's share (3.0-6.0%) is the part that travels between GPUs
while its milliseconds are the machine's, and the missing pitch-demo piece is a window, not an upload path.
docs: OQ-51 landed - wgpu on main, and the chain of checks that caught each other on the way
docs: the device clause is now a register, and the front page states its count
Mirrors docs/11 and README: gate P1.5's four targets live in docs/evidence.json as
data, the falsification table says which rules were proven by violating them, and
Home carries the derived 0 of 4 rather than a sentence nobody checks.
docs: front page claimed no culling, no WGSL and no compaction; all three exist and are measured
Reading the panel end to end rather than grepping it found more than the pattern search had: the
phase status, the open-question count, three capabilities listed as missing that have since landed,
a fourth-device phrasing that predates the gate naming four machines, and a WASM parity figure
quoted as current when it describes an older 13-line boot. The block is 18 lines and hashes to
db71f8a0... today, which the evidence manifest publishes and CI reproduces.
The standing lesson, and the reason for the read-instead-of-grep: docs that agree with other docs are
not verified. This page was internally consistent and publicly wrong.
docs: the public front page said 118 tests, 48 questions and no spikes run; all three are now wrong no longer
wiki: the WASM parity is a three-architecture matrix, and arm64 V8 matches its own native build
Run 38270620391: `parity ok on arm64` from macOS, `parity ok on x86_64` from Linux and Windows, 13 lines
identical each. Stated as what it is - the instruction set in the gate's M1-browser row, executed by V8
rather than a browser - because a green tick here is otherwise easy to read as four of four.
Also repaired a sentence the previous edit had broken mid-clause ("...every CI architecture and, runs
under..."): a wiki page with a dangling conjunction is the same class of defect as a dead link.
wiki: the roots run inside a WASM engine, and the blocker we cited for it was a misreading
docs/11's claim was that WASM execution needs the export-lint decision because a runner would have to be
wit-bindgen output. Both premises are true and the conclusion is not: a WASI command exports _start from
the Rust runtime, so nothing in this repository writes an export or an unsafe. The engine now builds for
wasm32-wasip1, runs in CI under Node's WASI, and diffs identical against the native build - 13 lines,
EPOCH_0 91b97a52... among them, main green in 38265695965. The step fails on empty output, because an
empty diff between two empty files is a green check that means nothing.
Home, Open-questions and the status log all repeated the old claim, so all three are corrected rather
than one of them quietly updated. Stated alongside it: this is a fourth kind of target, not spike S3's
fourth device - M1-browser still means a browser on an M1 - and the marched f32 values riding in the same
lines are presentation, not evidence (ADR-017). OQ-41 stays open for mods exporting functions; it was
never what blocked this.
wiki: the field becomes a picture on a CPU, with the first measured frame-cost shape
Mirrors 237c916. render::raster marches a pinhole grid and reports 45.5 evaluations per pixel, flat across
three resolutions on one tree, with the 33.0% of rays that find nothing charged 69.6% of all evaluations -
ADR-024's argument as a measurement instead of a derivation. The page states plainly that this is not R1:
no GPU, no WGSL, no BVH, and a normal-mapped colour that is not a material model because OQ-10 has not
settled lighting. Counts moved to 118 in twenty suites, run 38249012218, seven jobs green on all three
runners.
wiki: 111 tests, and the second header promise the loader was not keeping (depth)
Mirrors 33fee95. contracts/sdf.fbs says depth is carried so the loader can refuse before it walks
anything; nothing compared it until now. Recorded with the explicit note that this is not gate AS5 -
compile-time budget rejection needs a per-tree budget the docs do not define.
wiki: 110 tests, the grammar-version refusal the loader was not keeping, and the CDN flake with its cause
Mirrors 40dab12 and 67019d2. doc 09 has always said grammar_version is refused at load; the field was
carried and read but never compared, so the refusal - before a single node is parsed, exact equality
because the accepted window is OQ-30's question - is now in the reader and in the status log. Plus the
harness note: a 504 while fetching the pinned cargo-deny binary reddened trunk with six jobs green, the
downloads now retry, and the retry is recorded as preventative rather than proven since no 504 has
travelled the new path.
wiki: the runtime reads the file - reader, oracle fixture, engine boots from disk; 109 tests
Mirrors b4d4128, f72692a and 19ff900 plus doc 09 and doc 11's rebuilt rows. The new section records the
reversed bar (a reader fed flatc's own bytes), the off-by-four bug the fixture caught, why field
presence comes from the vtable, the five refusals, the falsified disk-boot assertion and the two checks
that saw different things (cargo test passing where clippy failed).
Home's blurb was four commits stale: it said 99 tests, no engine runtime, and A-02 among the unwritten
artifacts. All three are now measured rather than recalled: 109 in CI run 38244520595, identical on the
three runners, and crates/engine is a binary that boots.
wiki: the engine boot is a check, twist and bend land (13 of 15 kinds), 99 tests
wiki: the engine boots and asserts the published vectors; assetc lowers, the export ban waits on a decision
wiki: crates/render lands the grammar on the CPU, 86 tests, and R5 gets its first measurement (with its f32 bound stated)
wiki: 75 tests, artifact A-07's grammar, and gate P1.7 named for the persistence half of OQ-47
wiki: corrected S2 numbers (52 ms scattered / 7 ms clustered per full slab), 68 tests, and gate P1.6's bisection walk
wiki: mirror the commit path, crates/memory, and spike S2's first numbers (OQ-49)
Phase 0's table goes to 61 tests with the two new suites named, the spike row records
that P1.2 is met (57.2 M appends/sec, p99 200 ns) and P1.3 is not (4.9 ms pass / 93 ms
commit for one full 64 KB slab region), and the still-unmet paragraph is rewritten: the
commit path landed the same day it was listed as missing, the insertion sort became an
octant partition, the pass became linear, and no digest changed. Open-questions gains
OQ-49 and corrects OQ-47's trigger — slabs append records now; what is still absent is
the projection into M_active. Home's status blurb follows.
wiki: mirror ADR-030's sign-off, A-10's freeze and crates/ledger's first roots
Five pages move. Decision-log gains ADR-030 with its rejected alternatives and the note
that ADR-028's "not settled by it" half is now settled; Roadmap-and-status's Phase 0
table records run 38181630859 — 51 tests on three runners, the fifteen A-10 vectors
byte-identical across them, cargo-deny's first real audit now that BLAKE3 is in the
graph, and the still-unmet list extended to the commit path the hasher does not do yet;
Home's status blurb; Glossary adds Epoch root, Tree shape and Intensional commitment;
Open-questions marks OQ-44 live and OQ-47's trigger partly reached.
wiki: mirror A-03/A-08 drafting and the ADR-026/028 sign-offs
Repo truth at 3934020: 48 OQ (OQ-45..48 registered), 32 tests green on all
three CI architectures, SPEC-04's hashing half locked, A-03 and A-08 drafted
under DRAFT headers, and ADR-026 (wgpu) plus ADR-028 (BLAKE3-256) promoted
from proposed to locked.
docs: mirror the ADR-029 sign-off, A-01's freeze and the licence
SPEC tally is 4 locked / 4 proposals / 1 open now that SPEC-01, SPEC-03 and
SPEC-09 are locked; SPEC-06 survives as doc 03's allocation choice. Home gains
a License section, OQ-37 and OQ-43 lose the licensing half of their scope,
A-03 and A-08 are recorded as unblocked, and the Phase 0 numbers move to
CI run 38150859744 with 242 doc links.
docs: mirror the Phase 0 landing, ADR-029 and OQ-44 into the wiki
The pages still said "not a line of engine code exists" and 43 open questions. First
code has landed and been measured on three architectures, so the status boards, the
SPEC-03/06/09 rows and the ADR count (29, including the still-unsigned ADR-029) now
match docs/. Also carries the earlier fidelity fixes: the rejected render-pass equation
stays visible next to its sanctioned reading, Rollback points at doc 10, and the VAD
abbreviation is gone because docs/14 does not define it.
docs: add architecture, vision, roadmap, decision log, open questions and glossary pages