A local autonomous agent plans and executes software work end to end, runs tests, and opens pull requests while humans validate outcomes afterward. Greater autonomy increases leverage, but it also makes failures rarer, harder to detect, and more consequential. section 1.4.1
Use autonomy with engineering discipline rather than treating speed as the primary benefit:
- Match agent speed with tighter verification, governance, and review. section 1.4.1
- Expect systemic misalignment: one configuration error can produce a coherent but wrong change across multiple services. section 1.4.1
- Treat an agent as an exposure of weaknesses in a human-paced delivery process, not as a repair for them. section 1.4.1
Figure 1.2 The autonomy-risk spectrum: each step grants more leverage but demands tighter verification, governance, and engineering discipline.

The autonomy-risk spectrum makes the tradeoff visible: each increase in leverage demands stronger verification, governance, and engineering discipline. section 1.4.1
