Superagency and ADHD

“You can just do things” is tech industry meme¹, shorthand for the idea that individual agency is powerful, available, and underutilized².

Agency is a nearly universal value in tech, perhaps one of the few. “You can just do things” elicits both starry-eyed optimism and reminders of the more disruptive tendencies of the industry (see this excellent conversation with Tamara Winter for a discussion). Cate Hall, a popular Substacker, even has a forthcoming book on becoming more agentic. And, of course, you can’t read about AI at all without encountering agency in all of its forms.
Inflection points in my life have been characterized by agency; doing those things that no one said to do or expected me to do, making audacious decisions seeing them through. It’s an immensely fulfilling feeling.

And yet, agency often feels tantalyzingly, frustratingly out of reach.

About 30 years ago, I was informed that my inattention and restlessness had a name: ADHD (then called ADD) but that I would likely grow out of it. I didn’t. Since then, a combination of medication, personal organization systems, and relentless habit-tweaking have enabled me to mostly overcome the worst tendencies of my ADHD. However, none of these things have ever been capable of defeating my brain’s most powerful ADHD-enabled habit: procrastination.

Fellow ADHD-havers will know that “procrastination” means something different for many of us than it means to the neurotypical population, something more deeply rooted, an immutable part of existence. It means something much closer to “inertia” or even “paralysis”.

You’re probably thinking this is overly dramatic, that everyone procastinates, and that you can just do things if you really want to. And, for most people, that’s true. Everyone gets stuck at some point in their lives, or develop bad habits that require a lot of effort to break, but ultimately people are able to. Most typical guides to overcoming procrastination follow some predictable formulae: set a pomodoro timer, break the task into small chunks, block time on your calendar, reward yourself for making progress, identify what emotion is underneath the procrastination, etc.

ADHD-driven inertia easily resists all of these tactics by increasing the activation energy for getting started to just beyond your current energy levels, regardless of your current energy levels. It’s not that you don’t want to get started, or that you don’t understand the costs of not getting started, it’s that getting started feels so incredibly difficult that you can do almost anything else but start. Just draw the rest of the owl, indeed.

The only thing that semi-reliably breaks this apart for me is an unmissable deadline (though I’ve still missed plenty) or someone else depending on me to deliver. Don’t get me wrong, I have a somewhat successful career, an amazing family, and I manage to live a (mostly) normal life. ADHD inertia isn’t a constant, it doesn’t materialize for every task or project. In fact, not even for most tasks. However, when it decides to emerge, it can take hours, days, even weeks to finally move past it. I’ve never found any tactics that I can independently implement that reliably work; i.e., exercising agency is not usually an option. I’d made my peace and had accepted this is just how my brain works, treating these periods of being stuck as “seasonality” for my executive function. My brain just has fertile and fallow periods. This doesn’t help me get unstuck, but it does make me have more self-compassion and not inflict too much damage from the second arrow.

Agentic AI changes this.

LLMs have totally transformed lots of things that I do, and I use them constantly, but they didn’t fundamentally alter my ability to overcome ADHD inertia. But now, this thing that has seemed as close to an immutable fact about myself as can be is melting away. And it’s thanks to Claude Code.

ADHD inertia comes with not a small amount of embarrassment and shame. Admitting that you can’t get started feels terrible. Claude Code almost completely solves this for me. I can open a new session, and give it the worst possible prompt, stream of consciousness style, without thinking too much about planning or success criteria or anything else. In a few minutes, I have something to react to, and I can move from getting started mode to improvement and iteration mode. I can even throw away the first session, it literally doesn’t matter. My brain has now overcome inertia and I can make progress on an idea.

That’s it, that’s all it takes. Claude has successfully reduced the cognitive and emotional cost of “getting started” to approximately zero.

Of course, I work in a code-driven profession, so generating code is often what getting started looks like for me. But we all live in a world in which a surprising number of problems can be turned into code. And, there are already tons of people talking about how they just use Claude Code as general-purpose agent where code isn’t the objective. We’re very, very close to widespread, cost-effective access to agents who can help you get started on all kinds of things.

I don’t think there’s anything intrinsic about Claude Code that brings this about, it’s the agency of Claude Code that is the magic. You could use an open source model (Kimi K2’s API is Claude Code compatible) or the Gemini CLI, etc. If your particular inertia is completion-based rather than initiation-based, you could use the same approach. The possibilities are really quite staggering, and I don’t say that lightly.

Isn’t this cheating? How do you learn anything? Isn’t the code low quality?

Note that I haven’t use the phrase “vibe coding” anywhere until now, despite the fact that agentic coding assistants are a primary use case for vibe coding. The difference in what I’m describing is that the goal is not to use agents to produce something, even if they end up doing so. The goal for me is to create the conditions where producing something is even possible.

Today’s AI discourse is mostly focused on productivity gains and whether they are real or illusory. Do AI tools help coders move faster? “Productivity” is generally shorthand for “units of ouput per time unit or person”. For personal productivity, though, this misses the counterfactuals, the entirely new output that could have existed but doesn’t for whatever reason (ADHD inertia in my case). Doing existing things faster is useful but less appealing for creation. Doing entirely new things that would otherwise be lost to inertia is much more compelling to me — increasing the amount of interesting and creative endeavors in the world is positive-sum.

A final note on objections that inevitably appear when discussions of mental health and AI intersect, especially in the US. It’s undeniably true that systemic barriers exist to health care, especially mental health care, and that using AI as a tool for mental health is a delicate topic. These discussions, though, often involve some version of the idea that these systemic challenges should be the focus, not developing technological workarounds. I think this is not only unnecessarily zero-sum, but it relegates people who suffer in many forms to a purgatory where they must wait for societal problems to be solved (good luck!) or risk the ire of others shaming them for just trying to improve their lives.

Both things can be true — we shouldn’t rely on technology companies to develop solutions to systemic social issues and we also shouldn’t expect suffering people to wait for problems that predate their birth to wait for the “right” solution to appear. Asking us to do so is truly denying us of our agency.

Footnotes

1. A Twitter thread investigating the origins of the phrase ↩

2. Sam Altman, OpenAI CEO, on the launch of ChatGPT ↩