Fables of the Future, Just Not Like That

As I hinted at in the last AI & Me, the biggest, newest model should not run every step of an agent workflow.The obvious reason is cost. Why pay two to ten times more than you have to? Theless obvious reason is time. Fable (and whatever follows it) is slower. Time,even at the speed of AI development, is still valuable.An expensive and slow model still belongs in your workflow, but you need to useit where it matters.Use the bigger, richer model up front when judgment matters most. Let it explore the repo. Let it read the important files. Let it think through the shape of the problem, the traps, the unknowns, and the safest implementation plan. Then hand that plan to smaller, cheaper models that can do the more mechanical work.For coding agents, this maps well:Use the strongest model to understand the codebase and write the plan.Use cheaper models to implement scoped tasks from that plan.Bring the stronger model back at the end to review the work, catch integration mistakes, and compare…