Without a standard, no delegation
Delegation only works when both sides share a standard for how the work is shaped.
Here is the idea the whole course hangs on. Delegation only works when both sides already share a standard for how the work is shaped. When you hand a task to a teammate or to an agent, you’re not handing over keystrokes — you’re handing over a structure the other side already recognises. “Build this as a command handler behind a repository port” is only an instruction if command handler and repository port mean the same precise, standard thing to both of you. That standard is the shared mental model.
And it is mostly technical, not domain. Yes, the words of the business have to be shared too. But the load-bearing layer is a standardised set of components and conventions — the command handler, the port, the aggregate, the shape of a test — that encode implementation detail in a form a human and an agent both comprehend and can reproduce. Because the shape is standard, the agent can produce it predictably and you can recognise and check it on sight. Because it’s standard, the work can pass back and forth without being re-explained every time.
This is the whole basis of delegation: without a shared standard, every handoff becomes a renegotiation, and you can’t delegate what you have to re-specify from scratch each time. Without standard, no delegation. So the shared mental model is not a nice-to-have — it is the literal interface between your product intent, your engineers, and the agents they orchestrate. Most of this course is, underneath, teaching that interface: the small, precise set of components and edges that humans and agents can both hold.