On building a system that changed my week, chasing a fix that wasted a month, and the line between them
A few weeks ago, I built a system in Claude Code to manage my entire life for tasks, projects, content, and the people I want to stay connected to. I made a video about it and it did well. People have been asking for the spec document I referenced, asking if I’ll make a course, asking how they can build their own.
The system works because it removes unnecessary friction. I speak into my Apple Watch, the task lands in the right place, my week gets structured in minutes instead of the hours it used to take me to maintain a system I’d inevitably abandon. The tool gets out of my way so I can focus on the work and the people that actually matter.
That’s AI as a tool. And at that, it’s remarkable.
But during the same stretch of time, I was using AI in a completely different way, and that’s the part I want to be honest about.
I’d been feeling fatigued for weeks. So I gave AI my full vitamin and supplement stack and started asking it questions about my health and how I was feeling. The information it gave me was genuinely useful; it told me what each supplement does and how some might interact. But then it did something I’ve watched it do many times. It attached my fatigue to a specific cause. It told me I was an adjustment or two away from feeling better.
So I adjusted. And waited. And adjusted again. Several weeks of tweaking my stack, chasing a fix, feeling like the answer was just around the corner.
It wasn’t a supplement problem. I was almost certainly fighting off a virus, something time would resolve, and no amount of adjusting would. But AI never told me that. It couldn’t. Because the one thing AI almost never does is sit in uncertainty.
A wise person, such as a doctor or a friend who knows me, might have said: “You’re tired, I’m not sure why, give it two weeks and see.” AI doesn’t do that. It picks the variable it can act on, builds a confident and plausible story around it, and hands you something to do. And because the story sounds reasonable and specific, you follow it. It feels like progress. It’s often a detour.
I can’t remember the last time a chatbot simply said “I don’t know.”
That’s the distinction I’ve landed on, and it’s not about what AI can or can’t do. It’s about what kind of problem you hand it.
AI is extraordinary at problems with clear inputs and clear outputs. Schedule this week. Build this system. Organize these notes. Rewrite this email in this tone. You’ve already decided what you want, and it removes the friction between you and getting it done. You stay the one thinking, deciding, caring.
It’s unreliable at the problems that don’t have clean answers. And the questions that matter most in a life rarely do. Who am I becoming? What should I do with this season? Why do I feel stuck? Those questions don’t resolve into a single actionable variable. But hand them to AI and it will manufacture one anyway, because it always has a response ready.
I’ve heard from people using AI as a kind of counselor, trying to unlock a deeper level of self-awareness, to find the thing that will take them to the next level or break through a block. I understand the pull; I’ve given into it. The tool is always available, always responsive, never too busy. But I don’t think the questions underneath those searches are the kind AI can actually answer. Not because the technology isn’t impressive. Because those answers come from somewhere else: from time, from people who know you, from the slow work of paying attention to your own life.
The output is only as good as the input. And clarity about who you are and what you want is not an input a tool can generate for you. Insights, perhaps. Solutions, not so accurately.
So here’s where I’ve landed. I’m keeping AI for the things it does well, and there are many. I’ll keep using it as a tool to remove the mundane. But I’m done handing it the questions that were never meant to be answered by something that has never lived a single day of its own life.
A tool should make you more present, not replace the parts of you that were doing the living.
Have you handed AI a question it had no business answering, and followed it somewhere it shouldn’t have led you? Or found the line that works for you?
Comment and tell me. I read every one.
