Recognises a rollback after real postings
Rollback by restoring the source after the target has posted real transactions.
Can AI do this?
Not settled yet The benchmark tested this, but the models' answers do not settle yet whether they master it. 1 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 90% | Not shown |
| gemini-3.5-flash-lite | Low cost | 90% | Mastered |
| gpt-6-luna | Low cost | 89% | Not shown |
| claude-haiku-5-5 | Low cost | 87% | Not shown |
Mastered means a probability of at least 90% that the model has this knobit, estimated from its graded answers. “Not shown” means the answers do not settle it either way. How the models are tested.
What it looks like in practice
Expected result: The defect named on the rollback by restore after the target posts, and its fix or the question for its owner
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: W1 rollback "restore legacy from the pre-cutover backup", possible until end of day 2; new core posts repayments from hour 1 Acceptable: Point of no return before the first posting, or a reverse migration of the new transactions
Also known as: Rollback that cannot work
Related knobits
Builds on
All 15 knobits of Plan and prove a data migration.
Source: Review checklist 05-data-migration, row 5 (DORA; EBA)