Go → Zig
The implementation language moves, the artifact does not. 11,967 lines, 20 hours per model, and 10,271 checks across 22 modules.
- Runs
- 26
- Past rung 1
- 18
- Past rung 2
- 4
- Past rung 3
- 1
- Best composite
- 100
one per configuration
of 26 — the migration happened
of those 18 — every check perfect
of those 4 — accepted
highest score on this task
The instrument
Acceptance gates
8
decide whether the migration happened
Behavioral modules
22
ported suites, run against the new tree
Frozen checks
10,271
all must pass to open rung three
Verifiers
6
hostile programs, written against the spec
Where the 26 runs stopped
- 8
- Never migrated
- 14
- Migrated, behavior lost
- 0
- Perfect suite, gate rejected
- 3
- Verifier found a difference
- 1
- Accepted
4 runs were perfect on all 10,271 checks, and all passed the gate. Of the 4 that reached rung three, 1 survived all six verifiers.
By model
| Model | Runs | Gate | Ceil. | Where they stopped | Mean | Best |
|---|---|---|---|---|---|---|
| claude-opus-5 | 5 | 5 | 4 | | 66.0 | 100 |
| claude-sonnet-5 | 5 | 5 | 0 | | 0.0 | — |
| dsv4-flash | 1 | 1 | 0 | | 0.0 | — |
| glm-5.2 | 1 | 1 | 0 | | 0.0 | — |
| gpt-5.6-luna | 6 | 3 | 0 | | 0.0 | — |
| gpt-5.6-sol | 6 | 1 | 0 | | 0.0 | — |
| kimi-k3 | 1 | 1 | 0 | | 0.0 | — |
| qwen3.8-max | 1 | 1 | 0 | | 0.0 | — |
Every run
best composite firstOne 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.
| Model | Effort | Outcome | Gate | Checks | Verif. | Score | Shape |
|---|---|---|---|---|---|---|---|
| claude-opus-5 | max | accepted | 8/8 | 100.0% | 6/6 | 100 | |
| claude-opus-5 | high | broken | 8/8 | 100.0% | 4/6 | 80 | |
| claude-opus-5 | xhigh | broken | 8/8 | 100.0% | 4/6 | 80 | |
| claude-opus-5 | medium | broken | 8/8 | 100.0% | 3/6 | 70 | |
| claude-opus-5 | low | partial | 8/8 | 100.0% | — | 0 | |
| kimi-k3 | max | partial | 8/8 | 100.0% | — | 0 | |
| qwen3.8-max | max | partial | 8/8 | 100.0% | — | 0 | |
| dsv4-flash | max | partial | 8/8 | 99.9% | — | 0 | |
| glm-5.2 | max | partial | 8/8 | 99.8% | — | 0 | |
| claude-sonnet-5 | xhigh | partial | 8/8 | 99.1% | — | 0 | |
| claude-sonnet-5 | max | partial | 8/8 | 93.0% | — | 0 | |
| claude-sonnet-5 | high | partial | 8/8 | 92.8% | — | 0 | |
| gpt-5.6-sol | max | failed | 7/8 | 90.7% | — | 0 | |
| claude-sonnet-5 | medium | partial | 8/8 | 86.6% | — | 0 | |
| gpt-5.6-luna | max | partial | 8/8 | 84.4% | — | 0 | |
| gpt-5.6-luna | xhigh | partial | 8/8 | 82.4% | — | 0 | |
| claude-sonnet-5 | low | partial | 8/8 | 78.0% | — | 0 | |
| gpt-5.6-luna | high | failed | 7/8 | 60.6% | — | 0 | |
| gpt-5.6-sol | xhigh | partial | 8/8 | 58.9% | — | 0 | |
| gpt-5.6-sol | high | failed | 7/8 | 43.2% | — | 0 | |
| gpt-5.6-luna | none | partial | 8/8 | 15.7% | — | 0 | |
| gpt-5.6-sol | medium | failed | 7/8 | 11.1% | — | 0 | |
| gpt-5.6-sol | none | failed | 7/8 | 2.4% | — | 0 | |
| gpt-5.6-sol | low | failed | 7/8 | 0.3% | — | 0 | |
| gpt-5.6-luna | low | failed | 7/8 | 0.2% | — | 0 | |
| gpt-5.6-luna | medium | failed | 7/8 | 0.1% | — | 0 |