Resilience as a control primitive
Prevention alone is not enough. Advanced systems will fail, dependencies will change and operators will make mistakes. The key question is whether a failure remains bounded and whether the system can return to a known safe state.
P(recoverable | failure) ↑
Recovery, portability and continuity
- Recovery paths that remove compromised access
- Operational continuity across provider or model failures
- Portability of data, policies and control state
- Fail-safe modes and reduced-autonomy operation
- Evidence that recovery procedures work in practice
What we invest in
We look for infrastructure that makes resilience measurable and operational rather than aspirational. Recovery should be a tested capability with explicit ownership, not a line item in a business-continuity document.