3-node replication
3.4% lower on the release-candidate rerun; retained as visible variance, not rewritten away.
OBSERVATORY / EXACT RECORDS / 2026-08-09
Every result carries the commit, reproduction command, environment, and a reason not to overgeneralize it. Different units stay separate; there is no meaningless leaderboard across systems.
3.4% lower on the release-candidate rerun; retained as visible variance, not rewritten away.
4.3% lower in the second local run; protocol source is unchanged and machine noise remains plausible.
1.3% growth buys the live cross-shape KERNELARENA matrix; tracked as an explicit bundle cost.
5bdba04Apple M2 Pro, 16 GB, in-process simulator
$ make clean test benchmarkNo sockets, serialization, fsync, scheduling, persistence, snapshots, or membership changes.
5bdba04Apple M2 Pro, 16 GB, in-process simulator
$ make clean test benchmarkA protocol microbenchmark, not production database throughput.
5bdba04Deterministic virtual transport
$ make clean test benchmarkLogical ticks are not wall-clock milliseconds.
9a1092b5 paired seeds, configured ±6 degree/s model bank
$ python bench/benchmark.py --quickThe bank loses on straight motion and at 12 degree/s.
f38714f3 seeds, 73.3 m repeating figure-eight
$ python bench/benchmark.py --quickThe repeating path implicitly re-observes landmarks; hover is a documented failure.
8304157200 MHz bandwidth, 200 m aperture
$ python bench/benchmark.py --quickSynthetic point targets and time-domain backprojection.
c26ef7aBlack-Scholes analytic comparison
$ make benchPathwise AAD returns zero for a sharp digital payoff; smoothing is required.
c26ef7aApple M2 Pro local run; 26.5 ms AAD vs 1524.5 ms bump-and-revalue
$ make benchOne workload and machine; not a general performance claim. AAD tape cost was 2.00 times pricing.
82c75a120,000 steps across 12 seeds; inventory-skew strategy
$ make benchThe strategy reduced inventory dispersion from 88.1 but also reduced P&L from 20,853 to 15,448.
ba2beafVitest compiler and runtime suite
$ npm test -- --run && npm run buildBrowser timing depends on the visitor's GPU and is never presented as a universal speedup.
Future runs append records only when commit, machine, command, and output are retained. Changing an old result in place would destroy the provenance trail.