The AI chaos beast in your head

This article is a continuation of the Is AI taking the joy out of coding, or am I giving it away? article I wrote three months ago. I wasn’t planning to write a follow-up so quickly, but recent thoughts and time-off have made me think. In June, I said I was giving the joy away. It turns out I was giving away my focus too.

Three months later…

Did anything change? AI did, for sure. I had accepted its presence a long time ago, and the way I work had changed with it. But I started feeling something different. Something I wasn’t aware of until I took the time-off and reflected on my own thoughts and processes. While I was on my bike for the last two weeks, my brain and thoughts finally slowed down. I started noticing patterns in my thinking, moments of clarity that were previously drowned out by the constant noise of ideas and tasks.

When I came back to work this week and started talking to a colleague, we quickly ended up in a chat about how AI drains our energy. That was exactly what I was thinking about when I was on my bike, but I thought this AI chaos in my head was something I was dealing with alone. I didn’t realize how common it actually is until I shared my thoughts.

Notes became agents

Before AI, I started my day by reviewing my notes and planning my tasks. When I knew what to do, I opened Visual Studio Code, and started writing code. When I was in the flow, I could write for hours, and when something in my head popped up, I wrote it down in a notebook. Notebooks were my go-to tool. When I physically wrote something down, it felt more concrete and less likely to be forgotten. My brain also panicked less than when I put it somewhere digital. I can’t really explain why, but that is what I learned over time by testing different methods.

Today, notes have become somewhat obsolete. I start my day still the same way, but when I need to code, I open an AI tool like GitHub Copilot or Claude Code and tell it what to start working on. This is where things change. I don’t like to wait, so I start thinking about other things. How can I improve it? When it’s done, it can do this next.

The wait is where it goes wrong. That’s the moment I start another agent to work on a different task or idea. In some cases I have 4-5 agents running simultaneously, each handling a different aspect of my work or thoughts. In the beginning, this was exciting, but looking back at it, I realized it is a constant context switching from task to task, from idea to idea, which drains my mental energy and focus.

This context switching makes you lose focus. A month ago, I had various agents running. At the end of the workday, I closed the lid of my laptop, went home. At home, I opened my laptop again, and started closing the applications and I spotted an agent waiting for my input. I totally forgot about spinning up that agent with that prompt. I had to retrace my steps and figure out why I prompted it in the first place.

So what I noticed is that there are no notes anymore, only agents, because it’s quicker to spin up an agent than to write it down. It is quick, but it comes with other side effects.

The side effects of spinning up another agent

I already mentioned context switching, but another side effect is the mental load and time you put into the new agent, while the agent on your original task might still be running. You start prompting more and more: “Add this to the app. Oh, don’t forget to do this or that.”

This constant interaction with multiple agents can lead to a fragmented attention span, making it difficult to focus deeply on any single task. That is where, for me at least, the mental draining becomes an issue. You want to deliver faster and deliver more, so you put yourself under continuous pressure.

AI made us developers move so fast that we run faster than we actually can.

It’s just one more prompt

Prompting can also be addictive. You get results so fast that you get a sort of rush to do more. Say I get an idea for a new app. The first prompt creates a good starting point. Then you ask to add this. Then that. Before you know it, you’re hours in. In my mind, that is a constant fight to keep up with the pace of my own ideas, but also to tell myself where to stop and just take a break.

The pressure from above

It is not only us as developers pushing ourselves, but there is pressure from above as well. I’ve seen this happen firsthand. Project managers, for instance, can also make use of AI to validate feature ideas.

They can just ask their AI: “what do you think of this feature?” and get instant feedback.

What I also noticed is that they can ask: “would this feature fit in the next release?”

That is where the pressure from above becomes tangible, when AI decides whether a feature fits. It creates an expectation for the project manager without actually knowing the full context or architectural impact of the decision. So, developers get an extra responsibility to manage both the technical realities and the expectations set by AI-driven decisions.

Your brain on autopilot

As developers, we only have a few really productive hours a day. That time when you are fully focused on the task and where you are able to write code without distractions. With AI, I’m starting to wonder if we made that window smaller.

My focus is constantly on the next thing. It went from a couple of hours to a fragmented state, where deep, uninterrupted work becomes increasingly rare. I saw this pattern change over time. A year ago, I would question AI a lot when it did changes to a project. I made sure what it did was to our standards and aligned with our architectural vision.

Today, these AI tools became so good because they learned how I work, what my typical next steps are, and know the project even better than I do at times. That is where I see my brain going on autopilot.

The previous outcome was good, so the next outcome will be too.

A while ago, I built a native macOS app in Swift. I don’t know Swift. I haven’t written a single line of it. Claude Code built it, opened the app, took screenshots, found its own bugs, and fixed them. I use that app every day. It went from idea to product in a couple of hours, but it didn’t make me a better developer. I didn’t learn anything from it. It just made me solve an issue I had for a long time and gave me a quick solution.

At that point, we start losing connection with our own critical thinking and decision-making processes. Research suggests this may weaken our ability to plan, reason, and solve problems.

So, action needs to be taken. If I see myself doing this, I tend to step back, start a review process, start questioning the AI’s suggestions, etc…

Is the way we work still human-centric?

While talking to my colleague, I started reflecting on how much our work has shifted towards managing AI tools rather than purely focusing on human-centric problem-solving. It made me question if the way we work is still designed around human strengths and limitations, or if we are increasingly bending ourselves to fit the capabilities of AI.

Is our typical 8-hour working day still structured around human cognitive rhythms?

I believe our brains are able to only handle a handful of tasks a day. With AI we went over that limit, but at what cost? We do a lot more, but lose ourselves in the process.

A month ago, I tried an experiment. I worked on a single task, and while the agent was working, I picked up my Switch and played a game. Just to give my brain a break and avoid falling into the autopilot trap. It helped in a way, as it allowed me to not think about what is next, but just relax. The problem that occurred was that when the agent finished its task, I needed a couple of seconds to reorient myself and get back into the flow of the work.

Another way could be to wait, but that gives me more stress. While I watch the agent work, my brain doesn’t rest. It starts planning the next prompt, the next improvement, the next task, and before I know it, I’m spinning up another agent anyway.

You’re not alone

For a while, I didn’t understand where some of the stress was coming from. I believed AI was helping me a lot, and for sure it does, but I didn’t realize how much mental energy it took from me.

The chat with my colleague helped me realize that I am not alone in feeling this way. Many of us are navigating the same challenges, trying to balance the benefits of AI with the cognitive load it imposes.

It isn’t just you and I either. Researchers from BCG even have a name for it: AI brain fry, the mental fatigue that comes from overseeing too many AI agents at once. Axios wrote about how AI agents are scrambling power users’ brains, and scientists are already looking at what long-term AI use does to our brains.

So, AI is the problem?

Not necessarily, it is how we all use it.

AI is not forcing me to start yet another agent. It doesn’t tell me to add that “one more thing.” That’s me. The tools just make it very easy to keep going, because there’s no friction left to stop me.

We need to be aware of our own AI usage patterns and set boundaries accordingly. Not only to protect our mental energy, but also to ensure that we keep our skills sharp and our focus intact.

My boundaries with AI

Not boundaries against AI, but around how I work with it. While I was on vacation, and while writing this article, I started thinking about my personal boundaries. This is what I came up with:

• Fewer parallel agents: Two or three sessions I can actually follow, instead of five I forget about.
• A “later” list: When a new idea pops up while an agent is working, I will start writing it down again to review and address it at a later time. That way, I hope to remain a bit more focused on the task at hand.
• Staying in the loop: Start to ask more what the agent did, why it did it, and take a more active role in guiding its actions.
• Real breaks: Not a game while I wait, since that left me disconnected. What those breaks will look like is still something I am figuring out. Those breaks are mostly a handful of minutes, so a bike ride or a walk would take too much time.
• A hard stop: When I notice “let me just do one more thing,” that should be my signal to close the laptop.

None of this is revolutionary. It’s the same advice we’ve had for years about focus and context switching. The difference is that before AI, the work set those boundaries for us. Now we have to set them ourselves.

Your turn

I’d love to hear your thoughts and experiences. Share them in the comments or reach out to me directly. Let’s learn from each other and find ways to work with AI without losing our minds.

• Do you see the same patterns in your own experience?
• How do you set boundaries with AI in your own work?