There is a joke about Richard Feynman’s approach to solving problems:
1. Write down the problem.
2. Think real hard.
3. Write down the solution.
People attribute it to Murray Gell-Mann. Of course, “think real hard” is easier advice to follow when you have Feynman’s knowledge of physics. You don’t usually come up with quantum electrodynamics just by sitting silently for an afternoon.
But lately I’ve noticed how quickly I try to skip that step.
When I get stuck, I open AI. Sometimes it becomes a good brainstorming session, and sometimes I get an answer much faster than I could on my own. I’ve spent a lot of time learning these tools, and I’m glad I did. But somewhere along the way, spending an hour thinking about a problem started to feel like something I have to justify.
Why would I sit here struggling (aka thinking) if I could ask AI?
There is no specific person putting this pressure on me. Nobody stands behind my chair asking why I haven’t prompted Claude yet. It is just this feeling that I should move faster, that I could already be working on the next thing. Even when I decide to solve something myself, afterward I can think: I could have done that in 30 minutes with AI. Why did I spend so long?
It was my decision. And I’m still arguing with myself about it.
But asking AI immediately doesn’t always save me time. Sometimes I spend a while trying to understand its answer, then more time figuring out how the components work together. Eventually I understand the flow of the feature well enough that I can almost one-shot it. Getting to that understanding took time, and I keep acting as if I should have been able to skip it.
Come on, why do I feel guilty about thinking?
I have this feeling that almost every problem is solvable. Yet I can postpone something for weeks while my brain builds walls around it. Sometimes sitting silently and figuring out the first step is enough to start. Now I reach for AI to get through that uncomfortable part, but I’ve found another way to procrastinate there too.
AI can think for 15 minutes and give me an answer. Then I have to sit down and understand it. It can’t put that understanding inside my head; that part is still 100% on me, at least for now. So I end up with lots of open chats waiting for me to read them properly and think about what they say. I asked for help because I didn’t want to feel stuck, and now I’m putting off understanding the help.
This bothers me most when I’m working on engineering problems. I’ve become better at using AI tools, but I feel less confident in my ability to solve hard problems myself. I guess that’s where my “engineering depression” comes from. I’m an engineer at heart, you know. That ability matters to me, and having an answer waiting in a chat doesn’t tell me whether I still have it.
I don’t have a concrete example that proves my skills have degraded. That’s an uncomfortable thing to admit in an essay about this. I remember how it felt to solve something hard, though the feeling isn’t as clear as it used to be. I guess I need to pick something difficult and see how far I can get on my own.
I’ve also noticed small gaps in my recall. A city I’ve visited, or the title of a book I read three months ago. I can remember an insight but forget the author’s name. Often I remember where to find it in Obsidian, or enough details to write a prompt and retrieve it. Useful, yeah. But it still feels strange when I can’t recall something myself.
I can’t tell whether AI has anything to do with those gaps. Forgetting a title also doesn’t tell me much about my ability to reason through a difficult problem. What I do notice is the shortcut I’ve built: as soon as I can’t remember something, I ask AI. I barely give myself time to try. I’m not sure I like that habit, and I’d like to leave more room for thinking on my side.
There is already some research on this. A 2025 study of 319 knowledge workers found that greater confidence in AI was associated with less reported critical thinking. It doesn’t prove that my engineering skills are getting worse. But it gives me another reason to pay attention to how I use it.
I see beauty in learning how to break a difficult problem down and understand a complex system. I’ll write more about those skills later. For now, I want to give myself time to practice them without spending that whole time calculating how much faster AI might have been.
Maybe I’ll start by writing a Snake game without AI, just out of curiosity. I haven’t tried it yet, so I don’t know how much I’ll struggle or what I’ll remember.
I’d like to find out. And if it takes me an afternoon, I’d like to let it take an afternoon. A whole day? That’s fine too.