When needed, let AI work in a way that fits the task better
The same AI does not have to work in only one way.
The role answers “who is doing the work”; the working style answers “how.” Switch when the task changes without rebuilding the role or restarting the project.
Some tasks are best handled quickly and directly. Others benefit from longer reasoning, sustained execution, tool use, or a more deliberate step-by-step flow. Even with the same role, the right way to work can change with the task.
Separate the role from the way it works
FinchStory separates those two ideas. The role is still who is doing the work; how it carries out the task can be loaded or switched as needed. What you see is a different way of working, not a requirement to understand the Runtime, protocol, or execution engine underneath.
Change the working style when the task calls for it
An engineering role might use a lighter approach while discussing a plan, then switch to a more execution-oriented style when it is time to edit code, work with files, or stay on a longer task. The role does not need to be recreated, and the project does not need to lose its thread.
You do not need to understand the underlying architecture for every task. The useful question is simpler: should this be faster, deeper, or should the AI actually start doing the work?
What changed