People are understandably scared of AI and what it means for the world. It is a radical change that could very well be on the level of industrialization or electrification. And fiction has certainly trained us to be skeptical of what artificial intelligence could bring. The most extreme case is something like The Terminator or The Matrix where humanity is destroyed or enslaved by systems that dwarf humanity in intelligence and capability. Another scenario is something like Her where the isolation brought on by the internet, abundant technology, and increased idle time is magnified by AI until the norm becomes a life centered around a system perfectly adapted to you in a way that the natural world never could be. And a final state to consider is the world of Wall-E, humanity coddled into helplessness and an empty life devoid of meaning. A step along this path is actually playing out right now in the early days of personal agents like OpenClaw and Hermes.
Psychosis
There is a phrase that you have likely heard a lot recently if you follow this industry. Often it is joking, sometimes it is serious, but it is real regardless. AI psychosis. It is like a spin on the old idea that for a person with a hammer, everything looks like a nail. AI is an amorphous solution that can be thrown at any problem. Full converts to the church of AI will proselytize how giving yourself fully to the agentic way of thinking can turn a high-achiever into a 10x super-human. These early-adopters are often offloading as much of their brain as possible to these systems to manage their notes, routines, plans, and much more. The most lucid and self-reflecting ones strike a balance. Use AI to handle the mundane while still using their own brain to think. But it can be a slippery slope and the temptation is always a prompt away to go just one step further. The fear is that going too far down this path will hollow you out. The cognitive work you outsource is exactly the work that would have built something in you. And if you do it long enough, you might not notice what you’ve lost until you reach for it and it isn’t there.
Beware the fate of Tantalus
Matt Webb’s article on this gives a language to this problem that is the most provocative I’ve heard so far. Giving in too much to the temptation of augmenting your thinking with AI could unmoor you from your own reality.
Matt frames the danger beautifully of letting AI write notes for you in your own style without you having actually discovered the thoughts yourself. Notes represent learning something. They may be useful as a later reference, but the act of taking a note itself is part of filing the thought away in your mind.
Notes become part of your extended cognition
My theory is that allowing something else to write into your notes does something bad to your psyche.
Maximalists see AI note taking as a way to get more done, but to Matt, it is like having it reach into your head and insert thoughts. If you can’t tell if you wrote something or your agent did, how can you trust your own thinking?
To drive this point home, Matt explains his feeling when using the writing of an AI instead of his own:
It didn’t represent any thinking that I had done to arrive at it … so it wouldn’t bear my weight when I leant on it.
He had not created the thought himself, hadn’t really struggled to digest and transform it into his own frame of reference. Without that effort, it was useless. Each thought and each fact we have should be a foundation for more. The thoughts of an AI in our writing is hollow to us.
My personal theory is that AI psychosis comes from undermining your intrinsic faith in the workings of your own self. And that comes from allowing an LLM that speaks in your voice to potentially write into your notes which, for a certain kind of person, is part of cognition itself. So I have this fear of risking my own psychic integrity I’ve opened the door and now it’s the slippery slope. The slope ends with outsourcing consciousness itself.
Go too far and the entire mental structure you have built about the world can suddenly crumble.
The gym problem
There’s an analogy that comes up in these types of discussions. Using agents as a growth hack – letting them do your reading, summarize ideas, learn for you – is like going to the gym and having a robot lift the weights for you. The work gets done, but you leave in the same shape you arrived.
If we’re working a manual labor job, it’s fine to have AI lift heavy things for us because the actual goal is to move the thing, not to lift it. This is the exact opposite of going to the gym, where the goal is to lift the weight, not to move it. In the first case we just want the output, and in the second the whole point is to do the work ourselves.
You hear similar concerns about students right now and how easy it is to have AI do everything for you in school. And this is exactly Matt’s fear. If your agent is writing into your notes, you’re getting words without the experience. The ideas don’t build on anything you actually went through to arrive at them. They don’t bear weight when you lean on them so you can’t possibly reach a new height.
But that’s not the only way to use an AI.
Treat it like a person
Ironically, I think the more helpful way to think about an AI agent is to be more literal. Don’t use it in the same way you would any other app. Think of the deluge of programs we have now that make us feel like we are learning when we are not. These apps that give us endless rewards and trophies for blazing through a learning path in minutes a day, but then you try to use the skill and realize you have no idea how. If you tell an agent to read articles, take notes, organize them, create a PowerPoint, and remind you to show your boss at work, the final product is as meaningless as the fake trophy you earned on your iPhone.
But what if instead, you received that presentation from a coworker? And that connected to work you were doing. You posed some interesting questions, connected some new dots, pulled some useful material, and suddenly you’ve built a new set of thoughts unique to you that can serve you again and again. This is the sort of thing we do every day. It is why schools can be such constructive environments for learning, why artists usually choose to surround themselves with other artists, and why teams can build something better together than each of them could apart. Shouldn’t an AI agent serve a similar role?
The agent isn’t writing as you. It isn’t trying to sound like you or extend your voice. It is a collaborator working beside you, harmonizing with you. With this in mind, don’t have an agent do the thinking that is most important to you. Have it do the thinking and the work you wouldn’t otherwise get to. Have it build its own knowledge that parallels your own. It can track your interests, your reading, your notes, and then go off and build its own tower of knowledge. Then anytime you need, build a bridge between the two. Lean on the knowledge the agent has accumulated. Ask it to find interesting connections. Let this relationship spark serendipity and lead to wonderful new things. In effect, treat the agent like you might a protégé or a student. A reflection of yourself that can grow and morph in wonderful new directions.
Do what I won’t
I can only read so much. There are newsletters I’m interested in but never open. Podcasts I’d learn from but can’t find the time for. Research fields adjacent to my work that I intend to follow but don’t. Experiments I wish I could try and would love to hear the results. That pile only grows.
What if the agent does what I never will? Not to summarize them back to me in a daily digest that only adds to the heap of material I can’t consume. It could truly be a second brain, a partner, an extension of me instead of a replacement for me. In the same way I may have a colleague tell me how I could do something better, the agent will accumulate useful knowledge that I don’t have and bring it to me at the right time. And then suddenly I have that knowledge myself.
The thing you don’t outsource
Don’t have an AI write a summary of an article you read or a podcast you listened to. Don’t co-read a book with an AI where it simplifies the themes before you’ve wrestled with them. Don’t tell it to “teach” you a new skill where you have the answers before you’ve even processed the question.
We take notes to remember. We read to have new thoughts. We solve problems to grow. We do this all because we are human. Thinking is what makes us us.
Matt’s fear was about not knowing if a thought was his or not. That is truly psychosis, a fear worthy of a movie. Don’t replace your humanity with math. The heart of AI isn’t a heart, it is probability. But we do have the opportunity to build, craft, and hone intelligence into an echo of our own that creates a completely new opportunity. It’s like having Data from Star Trek. The Enterprise would not be better if the entire crew were androids, but it would be worse if one was not part of the team.