How AI Built My Burnout (And How I’m Fixing It) [Raw Session]

0:00 — AI Fatigue Feels Different

Hey, welcome back to Slow Builds.

I’ve talked before about AI fatigue, and I still think that’s real.

There are days when I’ve been using AI for hours. I’m going through code, reading responses, testing things, jumping between chats, and eventually my brain just gets tired.

That kind of tired makes sense.

Take a break.

Go to sleep.

Come back at it tomorrow.

Usually that’s fine.

But lately I’ve realized what I’ve been feeling is different.

I don’t think I’m just tired from using AI.

I think I burned myself out on AI a little bit.

And I don’t mean I’m done with it.

I’m not done with AI at all.

We can’t be.

It’s part of our lives now. It’s part of everything we do.

I still think it’s one of the biggest changes I’ve seen for building software, changing processes, and just being part of normal life.

There is value in it.

It’s not going away.

It’s here to stay, so we all have to learn it and learn how to use it.

But I also think I pushed it too far.

Not in one dramatic way.

More like slowly, one extra prompt at a time.

One more tool.

Another workflow.

Another idea.

Starting another project.

Eventually I got to a point where opening some things back up started to feel heavy before I even did anything.

That’s when I realized this was not just fatigue anymore.

It was burnout.

2:02 — I Went Too Deep

The first thing that happened is I went too deep.

When AI started moving this fast, and when everyone started talking about it, I felt left behind.

I’ve done videos on that.

I didn’t want to be left behind.

I didn’t just want to use the tools.

I wanted to understand them.

I wanted to understand the models, the agents, the workflows, coding with it, proper prompts, context, instructions, memory, all of it.

And all of it was useful.

It helped me see AI as more than just a chat box.

Once you start looking at tools like n8n, Pipedream, or Gumloop, you realize a lot of this really is workflow design.

Something comes in.

AI does one or two parts here and there.

It gives you responses.

You make decisions.

Then something gets pushed out.

That changed how I thought about what was possible.

But then there was always another layer.

Once I understood one thing, there was another tool.

Another thing to learn.

Another thing to figure out.

Once I understood that tool, there was another model.

Now I had to figure out which model worked best for which operation or task.

Then another coding agent comes out.

Then another way to structure instructions.

Another way to change prompts.

You can see where that goes.

You start going real deep.

And it always feels like, “I should probably understand that.”

That is a dangerous thought, because there is always going to be something else.

There are always changes.

Everything is moving forward.

And if you think you have to stay at the forefront of it, watch every video, read every doc, and test every new tool, it never ends.

So it has to reach a point where I can say:

I know enough to build with what I have right now.

I can learn the next thing when I actually need to.

4:13 — I Went Too Wide

After going deep, then I had the problem of going way too wide.

And this might actually be the bigger problem for me.

AI removed a lot of friction between having an idea and turning it into something.

Before AI, an idea had to survive a little bit.

I had to think about it.

Google it.

Write it down.

See what was out there.

Most ideas would probably just fade away.

And honestly, that was probably healthy.

They might still exist somewhere in a notebook, or an email, or somewhere in my brain, but a lot of them just died.

Because if I wanted to take an idea seriously, I had to spend real energy on it.

Now that first step is almost free.

The speed to research something, figure out an MVP, or see what exists is almost instant.

I can have an idea and immediately start researching it.

I can look at domains.

I can think through the business model.

I can see what already exists.

I can think through what might be legal or not.

And because AI will keep going with me, that idea starts becoming real before I decide whether it deserves to be real.

That can happen with software.

But it can also happen with a book, a product, a physical business, a game, a video, almost anything.

AI can help put the idea into reality.

Everything has somewhere to go now.

That sounds exciting, and it is extremely exciting.

But it also means fewer ideas naturally die now.

And I think I needed some ideas to die.

Not because they were all bad.

Just because I’m one person.

I didn’t sit down and decide I wanted twenty things going on at once.

I just stopped letting every idea end.

Eventually each idea had its own little world around it.

A chat.

A name.

A folder.

Research notes.

Sometimes a domain.

Sometimes a prototype already built.

Maybe a half-finished thing sitting somewhere that I now feel responsible for.

That’s where the width of the whole thing becomes exhausting.

It’s not just the number of projects.

It’s the number of open loops.

They are sitting there waiting for me to do the next prompt, push the next button, and move them forward a little bit.

6:42 — AI Creates Work Too

The part I didn’t really understand at first is that AI does not only remove work.

It also creates work.

A friend said to me the other day, “If AI is supposed to help us work faster and better, why does it feel like I’m doing more work than I ever did before?”

And sometimes that work is worth it.

But it is still work.

Especially with code.

AI can build something faster than I can build confidence in it.

It can write a feature in a few minutes, or even a few seconds.

But then I have to read it.

Understand it.

Test it.

See what else it changed that it should not have changed.

Make sure my tests are actually working.

Make sure the tests are built right.

Make sure it didn’t build the test for the new feature and ignore the old stuff.

If people are going to put their information into something I built, I can’t just trust it because the code looks convincing.

That review process takes a lot of energy.

And then if I ask AI to fix one issue, it might fix that issue and touch other things.

So now I’m reviewing everything all over again.

Checking everything again.

Testing everything again.

Trying to remember what I had already signed off on in my head.

That loop can go on forever.

It never ends.

Writing has its own version of this too.

AI can rewrite a chapter forever.

The book can be constantly rewritten, changed, fixed, rewritten again, changed again.

It can suggest structure changes forever.

It can make the writing smoother.

But then I have to check every single thing it says and make sure it still sounds like me.

Are the timelines right?

Is the meaning still what I wanted?

Did it change something it should not have changed?

Did it give me five more kids that I didn’t know I had?

That is work.

That is a lot of work.

It can produce a lot of content quickly, but it is a lot of work for me to make sure it lines up with my vision, my words, and my truth.

It is a different kind of work, but it is still work.

And after a while, the thing that was supposed to make everything easier started making everything feel like there was never an end.

8:51 — The Real Problem

I think the real problem is that I was letting AI decide how much work existed.

And according to AI, there is always more.

There is always another improvement.

Another feature.

Another tool.

Another test.

Another idea to throw in there.

Another thing I should probably look into.

AI does not naturally stop and say, “That’s enough for your life right now.”

It doesn’t know I have a job.

It doesn’t know I have a family.

It doesn’t know I need to work out, sleep, eat, and still have some kind of normal life outside of building things.

It can keep going.

I can’t.

And for a while, I think I was trying to work at its speed.

That is the part that burned me out.

Not AI itself.

Trying to live at the pace AI makes possible.

Because once you’ve seen that speed, normal progress starts to feel slow.

You can have an idea in the morning and something running by night.

That is extremely exciting.

But it can also mess with your sense of what a reasonable pace actually is.

Suddenly taking a week to do something carefully feels extremely slow.

Leaving a feature out feels like you are falling behind.

Letting a project sit feels like failure.

And I have a few sitting around, and it really does feel like failure.

But maybe it isn’t.

Maybe that is just what it looks like to build at a human pace again.

10:09 — Slowing Down

So I’m trying to slow down.

I don’t have a perfect system for this.

This isn’t me saying I solved AI burnout.

I definitely haven’t figured that out.

But right now it means being more intentional.

What am I actually working on?

What needs to happen next?

What is enough for this version?

And when am I done?

That last question matters more than I realized, because AI is not going to tell me when I’m done.

I have to decide that for myself.

Some projects might sit for a while.

Some features might never get put in.

Some ideas probably should stay as ideas.

Some chats should just end as chats.

And that’s hard for me, because I like building.

I love learning.

I like seeing what’s possible.

But these are all side projects.

They matter to me, but they are not worth burning myself out so badly that I have nothing left for the rest of my life.

AI did not remove my limits.

It just made them harder to see.

My time is still limited.

My attention is limited.

I have limited energy.

My ability to understand and trust what I’m building is still limited.

So maybe the answer, at least for me right now, is not to stop using AI.

It’s to stop letting AI set the pace.

I’m still building.

I’m still learning.

I’m still using it.

I’m just doing it slower.

And right now, slower feels a lot more sustainable than trying to keep up with something that never has to stop.

Thanks for watching. Hope you liked it.