The waves
If you follow the people who build with AI every day, you can see a strange rhythm. Every few weeks, the field finds a new center. First everything was prompting. Then prompting was declared dead, and everything was context. Then context was assumed, and everything was loops. Now the loop is losing its shine, and the graph is arriving. Some of the same voices that recently praised the loop now question it.
A way of working is announced, mastered, disputed and replaced in the time it takes most companies to schedule a meeting.
The water
I stopped trying to catch every wave a long time ago. I wait for mine. And the longer I watch the water, the clearer one thing becomes: I am already in it. Being there is not the same as riding everything that comes. I watch first: where a wave builds, who it carries, where it breaks. Most waves pass. Letting them pass is not the same as falling behind.
This is not distance from the field. It is the same posture I use in my work: observe first, and act on what is actually there, not on what is loudest.
A wave is a signal. It is not yet direction.
The primitives
Now look at the list of waves again, slowly.
Prompting is instruction: shaping what the machine is asked to do. Context is state: what the machine can see and carry while it works. The loop is iteration: act, inspect, correct, repeat. The graph is topology: which pieces exist, how they connect, and what can happen next.
None of the underlying forms is new. They belong to some of the oldest ideas in computing and systems design, surfacing again under new names. A wave often gathers around the bottleneck of the moment and settles when the bottleneck moves on.
And here is the part the commentary usually misses.
The hype gets refuted. The primitive gets absorbed.
The people questioning the loop today will still rely on iteration tomorrow. What dies is the fashion. What remains becomes standard practice, so standard that it turns invisible.
The spiral
This has happened before, and not only in software. In a program, a loop repeats instructions. In a control system, feedback changes the next action. In an organization, learning should change the next decision. In an agent, the same form now guides a machine that can interpret, act, inspect, and revise. The primitive has returned at another level of capability.
That is why this churn does not look like noise to me. When the lesson survives the fashion, the next cycle does not begin from the same place.
A loop becomes a spiral only when learning changes the next turn.
I have written elsewhere that intelligence is a loop. That essay was not written for the loop wave. It was written because the loop is the shape of how organizations learn, and that shape was there long before today's tools. This is why the primitives keep returning. They are not the fashion of the tools. They are the shape of the work.
The owner's question
For an owner, the lesson is not to study computer science. It is simpler than that.
Another wave will arrive soon. When it does, do not ask whether you are missing it. Ask which primitive it carries, and whether that primitive lands on a break that is actually yours. Most waves will not. The one that does is worth your full attention.
Building capability is becoming abundant.
Knowing which wave is yours, and which are just weather, is the part that stays rare.
One thought remains. The set of primitives is small, and it does not grow with fashion. The names of the next waves are still unknown. Much of what they will carry is not.
That is worth sitting with.