Interfaces for intention
For much of the modern era, many domains tied expertise to their machinery.
To make meaningful contributions, you had to navigate layers of accumulated friction: the workflows of finance, the toolchains of engineering, the protocols of healthcare, the planning systems of logistics. Using the system came before fully grasping the abstraction. The interface acted as a gate.
AI shifts that balance.
It doesn’t eliminate mastery, but it moves the bottleneck. As the cost of turning intent into execution tends toward zero, the interface stops absorbing most of the difficulty; intention begins to matter more.
When procedural fluency is no longer the primary constraint, the scarce part of expertise becomes the ability to articulate what should happen, under which constraints, and with what structural awareness. Skill gradually becomes less about technique and more about abstraction.
This raises a natural question: if expertise becomes intention shaped by constraints, what interface allows that to be expressed?
The first temptation is to assume chat becomes the universal interface. Chat was the first expressive surface AI unlocked. And chat is excellent for exploration: it lowers inhibition, widens the search space, and makes complex systems feel approachable. But chat is poor at governance; it captures desire but not the boundaries that give intention its form. It has breadth but lacks structure.
A world where intent moves in lockstep with execution requires a surface that makes structure visible; where constraints are explicit, flows are legible, and consequences can be understood before anything moves. A surface that reveals the architecture of what you’re trying to do, and accelerates the feedback loops between intention and outcome, before execution is locked.
If such a class of interfaces emerges, something downstream begins to shift. The borders of domains open up. When machinery is no longer the barrier, the transferable part of expertise becomes a person’s mental models: the patterns they recognize, the flows they can map, the constraints they can design. The penalty for stepping into adjacent fields drops, and recombination accelerates. Disciplines begin to overlap because their primitives can mix with less resistance.
We’ve seen this pattern before. The invention of the spreadsheet in the 1980s collapsed the machinery of finance and planning into a surface where structure became legible and malleable. People who had never touched a mainframe could suddenly model businesses, run scenarios, and reason in ways that had previously been locked behind specialized tooling. Marketers built forecasts without analysts. Operators ran capacity plans without dedicated systems. Founders did their own finance. Product teams modeled growth loops. The result was cross-pollination. Reasoning developed in one corner of an organization could now travel to another. Domains didn’t disappear, but their borders thinned. And the creators of that interface captured a meaningful share of the value unlocked.
All this suggests that one of the frontiers ahead of us lies in creating the interfaces that make this new kind of reasoning possible; the meta-interfaces that allow systems, people, and disciplines to combine in ways the old machinery never permitted.
Whoever builds these surfaces will not only unlock the value of recombination, but capture a part of it.