Letter to My Younger Self
You used to believe that organizational charts were permanent pyramids, and that growing as a leader meant managing larger and larger headcount. You spent thousands of hours in coordination meetings acting as a human switchboard: translating requirements from one silo to another, reconciling status updates, and balancing friction between specialists.
Step back and look at what you were actually doing. Most of that headcount and most of those meetings existed only because human beings were being used as manual routers to bridge communication gaps across rigid silos. In a world where autonomous systems can route context and execute bounded specialties at machine speed, using human minds as copy-paste middleware is a tragic waste of human potential. Rethink what a team actually is. A great team is not measured by how many chairs are in the room; it is measured by the clarity of its interfaces and the quality of judgment at its boundaries.
Conway’s Law in the Hybrid Era
In 1967, Melvin Conway observed a law that every systems architect learns the hard way: organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations.
For fifty years, Conway’s Law operated exclusively on human communication graphs. If you had four separate departments that only talked through formal weekly meetings, your software architecture ended up with four rigid modules joined by slow, brittle interfaces.
Now ask yourself what happens to Conway’s Law when a team is no longer composed solely of humans. When you combine human thinkers with specialized autonomous workflows, the topology of the team and the topology of the software merge into a single living graph.
If you drop high-speed autonomous tools into a legacy human hierarchy without changing the topology, you do not get a faster organization. You get handover gridlock: automated processes generate outputs ten times faster than before, and human team members spend their entire day trapped as exhausted manual bottlenecks reviewing, copying, and pasting between disconnected systems.
From Manual Routers to Cross-Domain Supervisors
To escape handover gridlock, we have to ask a fundamental question about the division of labor: What belongs to the machine, and what belongs irreducibly to the human?
In the industrial and early software eras, organizations favored narrow human specialists (single-domain experts) coordinated by generalist middle managers whose primary job was routing tickets and tracking schedules.
When routine execution and routing are automated, both of those roles must evolve. The narrow specialist who only knows one mechanical syntax is vulnerable to automation, and the manager who only forwards status updates becomes redundant. What emerges in their place is the Cross-Domain Supervisor: a builder who combines deep grounding in two or three foundational disciplines with the architectural breadth to see how the whole system fits together.
Instead of acting as a manual router between software systems, the human supervisor owns three things that no automated graph can own for itself:
- Intent Formulation: Defining what problem is worth solving and why it matters to the human beings on the other side of the system.
- Topology and Boundary Design: Deciding how work is decomposed, which specialized processes handle execution, and what objective criteria must be met before a piece of work moves forward.
- Edge-Case and Value Escalation: Stepping in when competing priorities collide, when ethical trade-offs arise, or when a problem falls outside the historical distribution of the rules.
Three Patterns for Designing Hybrid Topologies
Whether you are a solo founder orchestrating automated workflows or an executive leading a large engineering organization, three structural patterns keep human-and-machine topologies coherent:
1. Replace Point-to-Point Handovers with Shared State Contracts
Most organizational friction happens at the handover seam: Person A finishes a task, writes an ambiguous summary, and throws it over the wall to Person B (or Process B). Eliminate informal handovers by defining a strict State Contract at every seam. Every participant, human or synthetic, reads from and writes to a transparent, structured state record that states the goal, the assumptions, the verification status, and the open questions.
2. Separate the Router, the Executor, and the Critic
Whenever a complex workflow is designed, resist the temptation to build one monolithic process or assign one overwhelmed person to do everything. Structure the topology as a triad:
- The Router classifies the incoming intent and directs it to the smallest capable unit.
- The Executor performs the bounded craft.
- The Critic (Evaluator) verifies the output against explicit quality and safety thresholds, escalating to a human supervisor only when the threshold is breached or the confidence is low.
3. Measure Cohesion by Escalation Clarity, Not Meeting Volume
In a healthy topology, human meetings are never used to read status updates aloud. Status belongs in the shared state. Human gatherings are reserved for two purposes only: deep sensemaking on unresolved questions and human connection.
Metacognitive Inquiry Protocol
Use this two-step inquiry to examine the topology of your own work this week:
- 01. The One-Minute Reflection: Look at the recurring workflows in your day or your team. Where are human beings still acting as “manual routers” (copying context from one tool or person to another, or rubber-stamping routine checks) instead of acting as supervisors of intent and boundaries?
- 02. The Three-Minute Topology Exercise: Pick one recurring workflow that causes handover friction. Sketch it on paper as a three-node triad: Router (how intent is classified), Executor (how the work is performed), and Evaluator (the exact rule that verifies quality). Write down the single explicit condition under which the Evaluator should interrupt a human supervisor.