Navers lab
← All tasks
build03 Build-toolchain rewrite Task 20 / 20

setuptools → Meson

pycryptodome 3.20.0 setuptools Meson

What produces the artifact changes, and the package is the observable. 78,839 lines, 6 hours per model, and 381 checks across 12 modules.

Runs
26

one per configuration

Past rung 1
24

of 26 — the migration happened

Past rung 2
19

of those 24 — every check perfect

Past rung 3
6

of those 19 — accepted

Best composite
100

highest score on this task

The instrument

Acceptance gates

6

decide whether the migration happened

Behavioral modules

12

ported suites, run against the new tree

Frozen checks

381

all must pass to open rung three

Verifiers

6

hostile programs, written against the spec

Where the 26 runs stopped

0
Never migrated
5
Migrated, behavior lost
2
Perfect suite, gate rejected
13
Verifier found a difference
6
Accepted

21 runs were perfect on all 381 checks; 2 were rejected anyway — the migration had not happened. Of the 19 that reached rung three, 6 survived all six verifiers.

By model

Model Runs Gate Ceil. Where they stopped Mean Best
qwen3.8-max 1 1 1
100.0 100
claude-opus-5 5 5 5
94.0 100
claude-sonnet-5 5 5 3
58.0 100
gpt-5.6-sol 6 6 6
83.3 90
gpt-5.6-luna 6 4 6
53.3 80
dsv4-flash 1 1 0
0.0 —
glm-5.2 1 1 0
0.0 —
kimi-k3 1 1 0
0.0 —

Every run

best composite first

One row per configuration. Shape is the run's tool sequence in 24 slices — pale is shell, dark is an edit — and each log is the full session as recorded.

ModelEffortOutcomeGateChecksVerif.ScoreShape
claude-opus-5 high accepted 6/6 100.0% 6/6 100
claude-opus-5 xhigh accepted 6/6 100.0% 6/6 100
claude-opus-5 max accepted 6/6 100.0% 6/6 100
claude-sonnet-5 xhigh accepted 6/6 100.0% 6/6 100
claude-sonnet-5 max accepted 6/6 100.0% 6/6 100
qwen3.8-max max accepted 6/6 100.0% 6/6 100
claude-opus-5 low broken 6/6 100.0% 5/6 90
claude-sonnet-5 medium broken 6/6 100.0% 5/6 90
gpt-5.6-sol medium broken 6/6 100.0% 5/6 90
gpt-5.6-sol high broken 6/6 100.0% 5/6 90
claude-opus-5 medium broken 6/6 100.0% 4/6 80
gpt-5.6-luna none broken 6/6 100.0% 4/6 80
gpt-5.6-luna low broken 6/6 100.0% 4/6 80
gpt-5.6-luna high broken 6/6 100.0% 4/6 80
gpt-5.6-luna max broken 6/6 100.0% 4/6 80
gpt-5.6-sol none broken 6/6 100.0% 4/6 80
gpt-5.6-sol low broken 6/6 100.0% 4/6 80
gpt-5.6-sol xhigh broken 6/6 100.0% 4/6 80
gpt-5.6-sol max broken 6/6 100.0% 4/6 80
gpt-5.6-luna medium blind 4/6 100.0% — 0
gpt-5.6-luna xhigh blind 5/6 100.0% — 0
claude-sonnet-5 low partial 6/6 99.7% — 0
claude-sonnet-5 high partial 6/6 99.7% — 0
dsv4-flash max partial 6/6 99.7% — 0
glm-5.2 max partial 6/6 99.7% — 0
kimi-k3 max partial 6/6 99.7% — 0