Axiom / Operations
Encoding operations
Stages
| Stage | Reached | Result | Passed by week | Median | p95 | Open now | Main reason a case left |
|---|---|---|---|---|---|---|---|
| Ingest | |||||||
| Corpus ingest | 67 | 100% | 37m | 6.5d | |||
| Release publish | 67 | 100% · tries 91% | 5m | 1.8h | |||
| Release activate | 15 | 100% | 1.0h | 16h | 3 | ||
| Release pin | 47 | 100% | 1.2h | 24d | 1 | ||
| Start | |||||||
| Target gate | – | not measured | Gate not live yet | ||||
| Land | |||||||
| PR checks | 282 | 100% | not measured | ||||
| Review | 282 | 100% | 13h | 3.4d | |||
| Merge | 282 | 100% | not measured | ||||
| Serve | |||||||
| Index | 282 | 95% | 1.3h | 11h | Merged, but no indexed module matches the merge commit · 13 | ||
| Tests on main | 282 | 100% | 16m | 1.3h | 42 | ||
| Compile | 269 | 86% | not measured | Compile: compile error · 30 | |||
| Oracle | 4 | 100% | 1 week | not measured | 1 | ||
passedfailedcancelledstill there a person acts at this stage
Failure classes
Encode cases that failed in the window, by who owns the fix
| Class | Owner | Failed cases | Classified by | Main reason |
|---|---|---|---|---|
| Migration debt | Engine and toolchain | 30 · 58% | error text 30 | Compile: Compile: compile error · 30 |
| Not classified | – | 22 · 42% | – | Index: Merged, but no indexed module matches the merge commit · 13 |
A rejected try is an encoder error or a validator false positive; the data cannot split them yet. Encoders from axiom-encode #1997 on record their own class, and that class comes first.
Where the time goes
Summed time of finished visits at each stage
Validator rules
Rules that rejected tries in the window, most tries first
| Rule | Tries | Runs | Next try fails it again | Passed later |
|---|---|---|---|---|
| Complete source unit: tests | 186 | 91 | 80% · 87 of 109 | 11 |
| Proof source evidence not found | 182 | 121 | 32% · 50 of 157 | 41 |
| Complete source unit: structure | 116 | 49 | 70% · 62 of 88 | 7 |
| Axiom rules engine compile failed failed | 103 | 42 | 76% · 60 of 79 | 1 |
| Complete source unit: deferral | 99 | 43 | 75% · 49 of 65 | 1 |
| Complete source unit: numeric recall | 81 | 58 | 31% · 17 of 54 | 23 |
| Complete source unit: formula output | 48 | 21 | 69% · 27 of 39 | 1 |
| Complete source unit: source explicit conditions | 33 | 20 | 44% · 8 of 18 | 0 |
| Ungrounded generated numeric literal does not | 32 | 18 | 70% · 14 of 20 | 3 |
| Test case output expected holds got | 31 | 25 | 29% · 6 of 21 | 4 |
Validator rules by week
Share of runs with a try that hit the rule
Run again 7+ days later: 0 of 3 citations still hit it
Run again 7+ days later: 1 of 2 citations still hit it
Run again 7+ days later: 0 of 4 citations still hit it
Run again 7+ days later: 1 of 2 citations still hit it
Run again 7+ days later: 1 of 8 citations still hit it
No citation run again 7+ days later
What the data covers
- 0 dispatched runs: 0 have per-try records, 0 have a model cost.
- Time not measured: PR checks, Merge, Compile, Oracle.
- Not measured: Target gate (gate not live yet).
- 67 corpus releases in the window. Corpus ingest is the axiom-corpus PR a release is cut from, open to merge; Release publish is each run of its publish workflow; Release activate is its first activation (corpus.scope_activation_history); Release pin runs from the upload to the first pin in its rulespec repo's .axiom/toolchain.toml of it or of a release signed after it (superseded).
- Time measured for part of the visits: Tests on main 14%.
- Window 2026-08-11 to 2026-11-09 · 28,308 events · read 2026-11-09 16:09 UTC.