The people closest to the problem now have the most leverage.
Domain experts...
They know the concepts, vocabulary, and edge cases that let them steer the system well.
They can tell when a result is polished but wrong, shallow, or not credible. The likelihood of it confidently hallucinating things increases the closer you reach the edge of what's well-known.
The agent can execute the steps faster; only you know what is meaningful.
Harvard physicist Matthew Schwartz used an agent to complete a theoretical physics calculation that would normally take months, in two weeks.
“Claude is impressively capable, but also sloppy enough that I found domain expertise essential for evaluating its accuracy.”
“This may be the most important paper I’ve ever written — not for the physics, but for the method. There is no going back.”
“AI is not doing end-to-end science yet. But this project proves that I could create a set of prompts that can get Claude to do frontier science. This wasn’t true three months ago.”