Around August, I hit a wall. I'd been using AI tools for every aspect of my work for six straight months. Claude for thinking, Cursor for coding, GPT-4 for testing, Perplexity for research, Midjourney for design, local models for sensitive work. I was using more AI tools than regular tools. And I was exhausted.

Not physically exhausted. Mentally drained in a way I hadn't experienced before. I want to talk about what happened because I don't think I'm the only one, and nobody seems to be discussing this.

The constant context tax

Every AI interaction requires you to provide context. Here's my codebase. Here's the bug. Here's what I've tried. Here's my tech stack. Here are my constraints. When you're making 50+ AI interactions per day, you spend an enormous amount of mental energy on context preparation. Deciding what to include, what to omit, how to frame the question, which model to use for this particular task.

This is cognitive overhead that didn't exist before. Previously, I'd just think about the problem and code the solution. Now there's a meta-layer: should I do this myself or ask AI? If AI, which tool? How do I frame this? Is the output good enough or should I re-prompt? Each of these decisions is small, but they compound across a full day into significant mental fatigue.

The evaluation burden

AI output needs to be evaluated. Every response, every code suggestion, every generated test. Is this correct? Is it good? Is it subtly wrong in a way I might miss? This constant evaluation mode is draining. It's like being a code reviewer for 8 hours straight, except the code comes at you faster because AI generates it faster than any human would.

I found myself reviewing AI output with less attention as the day went on. By afternoon, I was accepting suggestions with cursory glances. This is dangerous. The bugs I caught in month one, I was letting through by month five. Not because I didn't care, but because my critical evaluation muscles were fatigued.

The FOMO spiral

A new model drops every week. A new tool launches every day. Each one promises to be faster, smarter, more capable. I was spending hours every week evaluating new tools, migrating workflows, and learning new interfaces. This is time I used to spend actually building things.

The fear of missing out on a productivity breakthrough kept me on the treadmill. What if the new Gemini update is actually better for my use case? What if this new coding assistant is the one that finally gets it right? The answer, almost always, is that the marginal improvement isn't worth the switching cost. But you don't know that until you've spent the time evaluating.

The skill atrophy worry

This one nagged at me constantly. If AI writes 60% of my code, am I getting worse at coding? When I need to write complex logic by hand, am I slower than I used to be? I don't have hard data, but the feeling was real. I'd sit down to write an algorithm and find myself reaching for the AI prompt instead of thinking it through myself. The muscle memory of problem-solving felt weaker.

How I recovered

I took a week where I didn't use any AI tools. Just me, VS Code (not Cursor), and my brain. The first two days were slow and frustrating. By day three, I was in flow states I hadn't experienced in months. Deep thinking without the interruption of "should I ask AI about this?" was refreshing in a way I didn't expect.

After that reset, I rebuilt my workflow with strict boundaries.

I reduced my AI tool stack to three. Claude for reasoning and review. Cursor for coding. Perplexity for research. Everything else got cut. No more tool-of-the-week evaluations. No more context-switching between six different AI interfaces.

I created AI-free blocks. The first two hours of my day are AI-free. I plan, think, and write code by hand. This protects my deep thinking ability and ensures I'm still exercising the muscles that make me a good engineer independent of AI tools.

I stopped using AI for things I enjoy doing. I like writing tests. I like debugging. These activities are satisfying and keep my skills sharp. Outsourcing them to AI saved time but removed joy from my workday. Time efficiency isn't everything.

I set a weekly tool evaluation budget. Maximum one hour per week for trying new AI tools or features. If it doesn't obviously improve my workflow in that hour, it gets discarded.

The balance I found

AI handles maybe 40% of my coding work now, down from the 60-70% I was trying to reach at my peak. But I'm more productive overall because I'm less mentally fatigued, I make better decisions about when to use AI versus when to think things through myself, and I'm not spending hours on tool management.

The sweet spot isn't maximum AI usage. It's sustainable AI usage. If you're feeling drained and you can't figure out why, consider that the constant cognitive overhead of managing AI interactions might be the cause. Take a break, reset, and rebuild with intentionality.