Infinite Starting

There is a specific kind of anxiety that does not look like anxiety.

You are working. Commits are landing, a new project is taking shape, the velocity would make a startup founder weep.

The thing you will not admit until 2am is that the project you should be working on sits in another tab, untouched, while your fingers keep typing into something that did not exist this morning.

Structured Procrastination

In 1995, Stanford philosophy professor John Perry wrote a short essay for The Chronicle of Higher Education titled "Structured Procrastination." It went viral before viral was a thing, eventually becoming a book, The Art of Procrastination: A Guide to Effective Dawdling, Lollygagging and Postponing (2012).

His core observation:

Procrastinators seldom do absolutely nothing; they do marginally useful things, like gardening or sharpening pencils or making a diagram of how they will reorganize their files when they get around to it. Why does the procrastinator do these things? Because they are a way of not doing something more important.

Perry's insight was that the to-do list itself is the mechanism. Put a daunting task at the top, the thing you are actually afraid of, and suddenly everything below it becomes urgent. You do tasks #2 through #20 to avoid task #1. The anxiety fuels productivity on everything except task #1.

He even proposed a hack: put a fake task at the top. Something that seems important but does not actually need doing. Now the real work becomes the procrastination activity. You do the real work to avoid the fake work.

Perry was writing about sharpening pencils. He did not anticipate what happens when the substitute task is building an entire product.

Displacement

Perry was describing what Freud had already mapped a century earlier. In his 1894 paper The Neuro-Psychoses of Defence, Freud introduced displacement: when the mind faces a threatening impulse it cannot process directly, it redirects the emotional charge onto a safer target.

Classic example: you are angry at your boss. You come home and kick the dog. The dog did nothing wrong, but the dog cannot fire you. The emotional charge moves from the dangerous target to the safe one.

His daughter Anna Freud systematized this in The Ego and the Mechanisms of Defence (1936). She placed displacement among the primary defense mechanisms, unconscious strategies the ego uses to protect itself from anxiety.

The same mechanism works on tasks. The anxiety is about the main project, the ambiguous one, the one where failure means something about you. Meanwhile, tweaking the color palette of a button is safe. It is bounded. You know when the colors look right. So your brain routes the anxiety there, and completing it gives you a genuine dopamine hit. Relief. Accomplishment. The loop reinforces itself.

Laziness would be easier to spot because you would be doing nothing at all. But you are working, genuinely working, and the exhaustion at the end of the day is real. That is what makes productive procrastination hard to see: you finished the day drained and cannot figure out why the original task still sits there untouched.

Judson Brewer, a psychiatrist and neuroscientist at Brown University, maps this as a standard habit loop in his book Unwinding Anxiety (2021):

Trigger: anxiety about the main project. Behavior: work on a safer, smaller task. Reward: relief plus the feeling of accomplishment.

Each cycle reinforces the next. The relief is real but temporary. The anxiety returns, often stronger, because now there is less time and the task still is not done.

The Supercharger

The old productive procrastination had natural limits. You can tweak button colors for an afternoon, maybe two. Then the colors are right, or at least right enough, and if you want to keep avoiding the main project you have to face it. The friction of the substitute task eventually runs out.

Agentic AI removed that friction.

Building an MVP used to require skills and resources that most people did not have. Now it requires a prompt. The cost of starting collapsed to near-zero. And the perverse outcome is that displacement now has infinite runway.

An afternoon on a color palette becomes weeks on a whole product. A product nobody asked for. A product that did not exist Monday and will not exist by next month because by then you will be building the next one.

The loop is the same: anxiety about the main project, redirect to a safer project, dopamine from shipping, anxiety returns, now with an extra unfinished project on the pile. But the velocity is different. Perry's procrastinator sharpened pencils. The AI-assisted procrastinator deploys to Vercel.

An AI coding tool can build a functional SaaS in an afternoon. Removing the single biggest barrier to starting is what the AI actually does. The blank page has always been the hardest part. Generating a working codebase from a prompt eliminates the paralysis that used to sit between having an idea and having something running. Once you have started, momentum carries you. The hours you put in are real, the exhaustion at the end of the day is legitimate. You are pouring genuine labor into something that exists solely because the alternative, facing the thing you ought to be building, felt too heavy to bear.

The Recursion

This is the point where the mechanism turns on itself.

Every new MVP adds to the backlog. The backlog itself becomes the anxiety source. Now you are not just anxious about the original project. You are anxious about the pile. A growing graveyard of half-finished repos, domains that go nowhere.

So you displace again.

You start something new because you are running from the weight of everything you already started.

Calling this a productivity problem misses the point. Productivity implies output is the goal. In this loop, output itself becomes the symptom. The commits land, the deploys go green. Every signal agrees that you are moving forward, and that is precisely what makes the loop so hard to break. The only thing you are moving toward is the next start.

Perry's hack, putting a fake task at the top, relied on the real work being more appealing than the fake work. But when the fake work is building something new and genuinely interesting, the hack breaks. The displacement target is now more appealing than the original task.

And So

In my Hyper-relevant Irrelevance post, I wrote about how algorithms optimize for engagement over exploration, trapping us in loops that feel like discovery but are actually convergence:

Perhaps the real tragedy isn't that we're trapped in these loops, but that we've become so comfortable in them.

The same dynamic is at work. The AI did not create displacement. Perry saw it in 1995. Freud saw it in 1894. For over a century the mechanism had a built-in speed limit: building anything real took time and skill. You could not displace your anxiety into a new SaaS because you could not build a SaaS in an afternoon. That friction was never designed to protect anyone from procrastination, it was just a side effect of how hard it was to make things. Agentic AI stripped that friction away. Now there is nothing between the impulse to start and the thing existing.

What happens when displacement has no speed limit?

I do not know. I keep thinking about the agents working through the night, building things nobody asked for, while the work you set out to do sits in another tab.