Series: The machine that never sleeps #

1. AI is making us work more
2. AI and the Devaluation of Effort (Current)
3. The myth of AI-powered Sisyphus

The value of a thing sometimes lies not in what one attains with it, but in what one pays for it - what it costs us.

• Friedrich Nietzsche

intro #

In Part I of this series, I explored the hidden paradox of modern, AI-fuelled work. That is, these tools that were supposed to free us are instead making us work more. The promised efficiency gains from automation have been subsumed by a never-ending need to just do more - not because we must, but because we can.

In this second part, I will examine a subtler and more insidious shift, not in how much we work, but in how we value what we do.

the moral weight of effort #

One of the quiet dignities of human labour lies in its cost. I have always found that I value things that demand something of me. Whether it’s a blog post I spent countless nights on, code that’s refactored over and over again, or projects that consume countless months of weekends and late nights. I believe that’s the stubborn human tendency to honour the investments we have already made.

Economists and psychologists have a name for this tendency to cling to our investments: the sunk-cost fallacy. In some way, they are right. Persistence can be irrational in strictly economic terms. But, I’ve never liked the idea of viewing human attachment as more or less a cognitive bug that needs to be patched.

I see the sunk-cost “fallacy” as recognising something fundamentally true; the finitude of human effort. We have only so much time and energy to do work. Once you have poured a portion of your life into a thing, no matter how flawed, finishing it becomes an act of respect for that time spent. The effort itself creates meaning.

And, generative AI unsettles this dynamic.

When effort becomes almost costless - when AI models can write thousands of words of prose on command, or generate working codebases from a paragraph - the relationship between investment and value breaks down. Beginnings proliferate and endings evaporate. The result is a new paradox that may be even more corrosive than the overwork of Part I:

Infinite capability breeds indifference.

The subtle shift #

2025 has been, for me, the year when LLMs truly crossed that threshold from being a fun, fascinating toy to mess around with, to a tool that can actually get stuff done (albeit, with some handholding). During that time, there has been a subtle shift in my appetite for work that I have only recently noticed.

Before, energy and time constraints enforced prioritisation. I had to choose: one side project at a time, one new feature instead of five, etc.. Scarcity was built into the system, and with that came the understanding that I just couldn’t do everything.

Now, with AI copilots and “agents” beside me, the world feels frighteningly frictionless. A new project can be mocked up in minutes. I can explore new frameworks in just a few hours with an agent by my side. And, when new curiosities arise, deep research tools can put together long, detailed dossiers within minutes.

So, I start everything.

A repo here, a new essay outline there. Because I can. The thing that once stopped me at the threshold - a daunting blank page - has been erased. And with every effortless start, the instinct to finish dulls just a little bit more.

from overwork to over-creation #

In this environment, what began as over-work metastasises into something even stranger: over-creation. A kind of manic productivity where every spark feels actionable, every passing curiosity becomes a half-started project, and where, paradoxically, none of them truly matter.

The problem isn’t that AI makes us work harder. It’s that it makes starting so easy that finishing loses its meaning.

I open my projects directory and see dozens of projects with one or two commits. I don’t even bother pushing them to my GitHub anymore. I know most of them are not going anywhere. My Obsidian is filled with outlines and drafts that are nothing more than shells of ideas waiting for me to breathe life into them. These are all just graveyards of beginnings, none polished enough to share.

“Anxiety is the dizziness of freedom” - Soren Kierkegaard

The abundance is paralyzing. When you can do anything, how do you choose what’s worth doing?

the psychology of frictionless creation #

What happens, then, is that effort and reward become decoupled: when the input feels trivial, the outcomes become disposable. This isn’t just about AI, it manifests itself similarly in instant streaming, once-click purchases, short-form content. But, within knowledge work, generative models accelerate this dynamic to an almost pathological degree.

Let’s consider three principles:

Low friction = low attachment: something that costs very little to begin, also costs basically nothing to abandon. I have projects that I laboured on for months that I just can’t let go of because they feel precious; whereas a project scaffold generated in seconds feels disposable. We don’t mourn what didn’t cost us.

Abundance dulls meaning: If scarcity creates value, then abundance destroys it. We can understand this when it comes to material goods, but we resist applying it to our creative output. Yet the logic holds: when everything is possible, completion loses its shine.

The scarcity that once gave meaning to effort has vanished. Where we once said I can’t do everything, we now whisper I can, so why not try?, then find ourselves adrift in a sea of half-started things, each one a monument to wasted potential energy.

the dark mirror of the productivity dream #

In the end, what I find ironic is that, the more AI amplifies our potential, the more it dilutes our motivation.

Effort, like currency, only has value when it’s limited. When dollars are infinite, we call it hyperinflation (something with which I have first-hand experience). When leverage is infinite, it devolves into a sort of creative nihilism, the quiet sense that nothing we make really matters because we didn’t really make it. At least, not in the way that counts.

AI removes the price tag from creation. It collapses the hours into seconds, the sweat into prompts. And in doing so, it makes the act of creation somewhat hollow. We become curators of machine output rather than authors of our own work. The satisfaction of having made something fades into the anti-climax of having asked for something.

This is the dark mirror of the productivity dream. We wanted tools that would free us to do more. We got them. But we forgot that the slow, frustrating, often tedious process of creation was where the meaning lived. Not just in the finished product, but in the making.

The easier things become, the less satisfying they feel.

what gets lost #

I started writing and coding long before generative AI. Back then, the work was hard. At times, it was maddening. And, it was also inefficient.

But, everything I did back then was also mine in a way that feels increasingly rare.

When AI scaffolds a project, plans the database schema, or refactors my code, it’s doing the parts I used to struggle through. And, somewhere along the line, something vital gets skipped. Not just the labour - I’m not here to champion working just for the sake of it - but the becoming that happens through labour. The way you don’t really understand an idea until you’ve tried and failed to explain it. The way debugging teaches you to think. The way revision solidifies what you believe.

Effort isn’t just a cost, it’s a catalyst. And when you remove it entirely, you don’t just save time. You lose transformation.

reclaiming the value of effort #

I am not advocating for rejecting AI. That ship has sailed. I also don’t want to sound like I am romanticising unnecessary suffering. The question isn’t whether to use these tools, but how to use them without losing ourselves in the process.

The only way meaning survives contact with automation is through friction by choice. Not the friction of bad tools or artificial gatekeeping, but deliberate constraints that create focus. This can be limiting yourself to a single project, working on a single essay, or learning one new tool, technology, or framework at a time. Let boundaries be real again.

Once you’ve picked that one thing, finish it. Completion itself is a discipline worth relearning. It doesn’t have to be polished or perfect. Just done. Published. Shipped. Closed. Then, to quote the cult of done manifesto, “once you’re done you can throw it away”.

In the end, while AI can multiply our work, it cannot multiply our care. It certainly can’t choose what matters. It doesn’t feel the weight of time invested. It has no sense of the satisfaction of turning something hard into something real.

the end in sight #

If Part I was about how AI made us work more, and Part II is about how it made us care less, then the last part will close the loop.

In Part III, I will examine the Sisyphean task of trying to do everything, of chasing every possibility in a world without friction, of saying yes to every idea because the cost seems so low, and of learning the quiet art of letting the boulder roll back down the hill.

Because maybe the real freedom AI offers isn’t the freedom to do everything.

Maybe, it’s the freedom to choose what’s worth doing, and let the rest fade away.