New tool helps AI write cloud infrastructure configs by asking clarifying questions instead of guessing

Researchers built a system that generates multiple possible cloud infrastructure configurations from vague human descriptions, then asks targeted questions to narrow down which one you actually want. This means AI can stop guessing at what engineers mean and start asking — reducing the back-and-forth that wastes time when you can't just test a configuration quickly.

Cloud infrastructure configuration is expensive to test wrong. You can't run a failed deployment cheaply or iterate fast, so AI systems that guess at ambiguous requests waste money and time. This approach uses a simple principle: break the problem into layers (what resources do you need, how should they connect, what properties should they have), identify where the AI's candidate answers disagree, and ask about the disagreements that matter most. The practical effect is moving from one-shot guessing to guided clarification — the same reason a good technician asks questions instead of making assumptions.