New AI Tool Lets You Choose Fast, Cheap, or Accurate — Instantly

When an AI does a big job — researching a topic, writing code, answering a customer question — it doesn't happen in one shot. It follows a chain of steps, often called an agentic workflow: an AI that can look things up, use tools, and take actions. Until now, the tools that design these chains have typically chased one goal: get the answer right. Or they blend all the goals into a single score, which locks in one fixed trade-off. If you later decided you cared more about speed than perfection, you'd have to rebuild the whole thing from scratch.

MoFlow, a new research paper from a six-author team, takes a different approach. Instead of committing to one balance, it maps out a whole menu of options at once — think of a travel site showing every flight from cheapest to fastest, rather than forcing you to rebook each time your priorities change. Once that map exists, you pick the trade-off you want and it hands you a ready-made workflow. No retraining, no waiting, no starting over.

The researchers tested MoFlow against six strong competing systems across six sets of tasks covering math, coding, and question answering. To keep it fair, they gave the rivals an advantage: those competitors got to retrain fresh for each new preference, while MoFlow never saw those preferences at all. Even so, MoFlow scored best overall on average.

The catch: this is a lab result, not a product you can use today. It was measured on academic tests, not real business messes, and nothing suggests it's shipping soon. But the underlying idea — one AI that adapts to your priorities instead of the other way around — is where the field is clearly heading.