Explains cutover and rollback
Rollback possible until the target's first real posting; go/no-go thresholds (DORA Art. 9).
Can AI do this?
AI does this well Most of the tested AI models master this. 4 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 100% | Mastered |
| gemini-3.5-flash-lite | Low cost | 99% | Mastered |
| gpt-6-luna | Low cost | 100% | Mastered |
| claude-haiku-5-5 | Low cost | 96% | Mastered |
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: Rollback possible until the target's first real posting; go/no-go thresholds (DORA Art. 9).
How it is assessed: Answers a written question on the cutover and rollback correctly
Related knobits
Needed for
- O-504 Recognises a cutover with no rollback
- O-505 Recognises a rollback after real postings
- O-506 Recognises go/no-go criteria with no thresholds
All 15 knobits of Plan and prove a data migration.
Source: Practice the stage 5 reference and review checklist rest on