From drift to ship

The smart ones are the ones I worry about.

They have done everything they were told to do. They subscribed to the right newsletters in 2023. They took the prompt-engineering course in 2024. They went to the AI conference in Singapore last October and the one in Bangkok in March. They know what retrieval-augmented generation is. They have used three different agent platforms. They post about AI thoughtfully on LinkedIn twice a week. They are, by every visible metric, in motion.

And they are tired. Quietly. Privately. The kind of tired that doesn’t make sense to them, because by all accounts they are doing everything correctly.

What they don’t realise is that they are not sailing. They are drifting. Fast, expensively, with the wind in their hair — but drifting. And the AI decade, the one they thought they were riding, is moving past them while they perform the rituals of keeping up.

The shape of drift

Drift in the AI decade has a particular look. It is not idleness. Idleness, at least, is honest. Drift is the opposite: enormous visible effort, all of it pointed at the next thing instead of into anything that compounds.

You will recognise it in your own week if you look. The hour spent reading about a new model that you will not use. The webinar attended out of vague obligation. The LinkedIn post crafted to demonstrate you are thinking about AI, not because you have something to say. The new platform signed up for, browsed for fifteen minutes, and forgotten. The course half-finished. The conference where you took notes you have not re-read.

None of this is wasted exactly. You have technically learned something. The problem is that none of it is yours specifically, and none of it is permanent. By the end of a year of drifting, you have absorbed enormous quantities of AI content and produced no artefact — nothing that exists in the world that wouldn’t exist if you had spent the year doing literally anything else.

That is the diagnostic test. After a year of drift, the only thing different about you is the number of tabs in your browser.

Why smart people are most at risk

The cruel mechanic of drift is that it preferentially catches the people best equipped to keep up. Smart, conscientious, high-agency professionals can absorb new tools faster, attend more events, take more courses, post more thoughtfully. They can run further on the treadmill before noticing they are on one.

People with less cognitive horsepower drop out earlier. They learn one tool, decide it is enough for now, and go back to doing their actual work. Their AI fluency is shallower, but their hands stay on the thing that compounds — their craft.

The smart drifter ends up further behind precisely because they had more energy to spend on motion. The energy they spent running fast in no particular direction is energy that did not go into building their moat, into encoding their trained instinct, into shipping their agent. A year of drift is a year not spent on the few things that could only be done by them.

That is the most expensive kind of year you can have in this decade.

What is missing is an anchor

Here is the part most diagnoses skip.

The opposite of drift is not more activity. More activity is what got you here. The opposite of drift is an anchor.

An anchor in the practical sense is three things, none of them mystical.

Stakes. The work you actually care about getting right — the kind whose failure mode you would feel personally, not just commercially. Stakes tell you which tools to pick up and which to put down, because the test stops being is this trendy and becomes does this serve the work I actually care about.

Standards. What good looks like, in your hands specifically. The bar you would not lower for a deadline. The kind of output you would refuse to ship under your name even if no one would notice. Standards turn AI from a magic shortcut back into a tool you supervise, and supervision is the part that compounds.

A direction. A small number of things you are building toward. Not a five-year plan — just a clear answer to the question what permanent artefact will exist by the end of this quarter that didn’t exist at the start? Direction collapses the option space. Most tools become irrelevant when you have a direction. That collapse is the point.

With those three in place, AI stops being a treadmill and starts being a wind. You still get the leverage. You also know which way you are pointed.

Your trained instinct is the moat. AI is the leverage. What you are rooted in — the stakes, the standards, the direction — is the anchor. Without an anchor, the leverage just spins faster.

The ship test

Three questions. Honest answers. They will tell you whether you are sailing or drifting.

One: at the end of last week, what permanent artefact existed that didn’t at the start? A document, a framework, a working agent, a system, a published essay, a redesigned process. Not a draft, not a thread of LinkedIn engagement, not a course in progress. A thing. If you cannot name one, you drifted.

Two: if you stopped reading new AI content for the next six months, what would you still be working on? If the answer is blank, you have been substituting consumption for work. If the answer is clear, you have an anchor. Most people don’t notice this distinction until they ask the question.

Three: when AI does the visible deliverable for you in half the time, what is the part you keep doing yourself anyway? That part is your anchor pointing at itself. It is the work that is yours specifically — the standards and stakes you are not willing to let the model take from you, even when it could. Build outward from that.

What changes when you anchor

The first thing that changes is your calendar. Most of the tool-evaluation noise stops feeling urgent. The conferences you used to feel obliged to attend become optional. The newsletters you used to skim become things you read deliberately or unsubscribe from. The cognitive bandwidth that was funding the treadmill comes back.

The second thing that changes is the shape of your output. You stop producing flat, fast, generic deliverables. You start shipping fewer things at higher density — documents, AI agents, frameworks — that are recognisably yours. Each artefact compounds the next. People begin to find your work the way they used to: by recommendation, not by algorithm.

The third thing is harder to name. The tiredness lifts. You stop feeling like you are on the wrong side of an arms race. The AI decade stops being something happening to you and starts being something you are inside, on your own terms.

What I want to leave you with is this:

Smart people are losing the AI decade not because they are not trying. They are trying very hard. They are losing because trying without an anchor produces drift, and drift in this decade is unusually expensive. The energy you spend running without direction is energy that doesn’t compound into anything that could only have been built by you.

The fix is not more activity. It is an anchor. Stakes, standards, a direction. From there, your trained instinct is the moat, AI is the leverage, and what you sail toward is yours.

If you don’t know where to start mapping that anchor, the essay on taking bearings is the practical companion. The full path lives at The Anchor Method. The cohort — here — is built around walking it with structure.

The treadmill is optional. Step off.