Transcript:

We’ve been noticing something over the past few weeks.

There are two very different ways people are beginning to approach AI, and they’re starting to diverge more clearly.

One path is focused on productivity and scale. The idea is simple: if AI can help you do more, faster, then you gain an advantage. You produce more content, build more systems, move more quickly than the people around you. The assumption underneath that approach is that more output creates more value, and that speed compounds into dominance.

That’s the world of 10x productivity. More workflows, more agents, more automation. And to be fair, there’s truth in it. AI does make people faster. It does expand what one person can produce. And in certain markets, that absolutely matters.

But there’s another path emerging alongside it.

And it’s quieter.

Instead of asking, “How do I do more?” it asks, “How do I think better?”

Instead of optimizing for output, it focuses on clarity. On judgment. On understanding what should be done in the first place, and how to know whether the result is actually correct.

Because in many domains—especially the ones built on trust, responsibility, and consequence—the constraint is not speed. It’s correctness. It’s whether the output can be relied on. Whether the reasoning holds. Whether the system behaves within acceptable bounds.

From that perspective, AI is not just a tool or a vending machine where you insert a prompt and receive an answer. It’s something that has to be shaped. Directed. Evaluated. Its usefulness depends less on how quickly it can generate something, and more on how clearly its role has been defined.

And that leads to a different kind of skill.

Not prompt writing.

Not tool stacking.

But the ability to define what success looks like. To anticipate where things can go wrong. To create boundaries, constraints, and evaluation criteria that exist outside the system itself.

This is where the idea of agents comes in.

There’s a growing recognition that simply giving people access to powerful models isn’t enough. Most people won’t know how to use that power effectively. So the interface is shifting—from one-off prompts to systems that can act, iterate, and maintain context over time.

But that shift doesn’t eliminate the need for human judgment. It actually increases it.

Because an agent can only operate within the structure it’s given.

If the goal is vague, it will pursue it vaguely.
If the constraints are unclear, it will cross boundaries.
If the definition of “good” is weak, it will produce something that looks right, but isn’t.

Agents don’t solve the problem of thinking. They amplify it.

Which means the real leverage isn’t just in doing more. It’s in defining better.

In being able to say:
What is this system supposed to do?
What counts as a correct result?
What are the failure modes?
How will we know if it’s working—or if it’s quietly drifting?

That’s a different orientation.

And it leads to a different kind of work.

Not building faster for the sake of speed, but structuring systems so they behave correctly over time. Not chasing output, but cultivating judgment.

Both paths will continue to develop. And both will have their place.

But they are not the same.

One scales activity.

The other sharpens thinking.

And the gap between those two is going to matter more than most people realize.