Most people ask a local model to do too much at once: understand the idea, clarify the requirements, design the architecture, write the code, test the work, fix the bugs, and decide what comes next. Then, when the result falls apart, they blame the model.

It changes what we should optimize, how we should architect agents, and where teams should invest their time. The harness is not “glue code” anymore. It is the control layer that decides what the model sees, what state persists, when work is decomposed, how verification happens, and when the run is allowed to stop. In practice, that means the harness is often the difference between an agent that looks clever in a demo and one that behaves reliably in production.