I caught myself doing it again last week. I was staring at an empty file, trying to figure out the right approach for a tricky refactoring problem. The kind of problem where you need to hold several moving pieces in your head at once and see how they connect. About three minutes in, my brain started reaching for the escape hatch. "Just describe it to the AI. Let the agent figure out the approach. You can review the output and iterate from there."
Three minutes. That's how long I lasted before the discomfort of not knowing became something I wanted to outsource.
The blank page escape hatch
I was listening to Cal Newport on the Modern Wisdom podcast recently, and he named something I'd been feeling but hadn't articulated. He argued that people are increasingly using AI to avoid cognitively demanding moments, not because the tool is bad, but because our brains are already so fried from context switching that the blank page feels unbearable.
Newport's data is striking: knowledge workers now switch to a different communication tool once every two minutes. The context switching is brutal and by the time you sit down to actually think about something, your brain is running on fumes.
Then AI offers the perfect escape. Instead of powering through those first painful minutes of loading a problem into your head, you offload it. The agent does the heavy lifting while you move on to the next thing.
Here's what I've started noticing, though. Those painful first minutes? That's not wasted time. That's your brain doing the most important work it does.
The strain is the signal
It takes 10 to 20 minutes for the human brain to fully load a complex, abstract problem. To inhibit all the unrelated circuits, activate the relevant context, and reach the state where ideas start flowing. Everyone knows this feeling: you sit down to write or design something, and the first 10 minutes are miserable. You're making no progress, everything feels forced, and you're convinced you have nothing useful to say. Then something shifts. The gears catch. And suddenly you're producing work that actually matters.
That initial strain isn't a bug. It's the cognitive equivalent of warming up before a workout. Skip it, and you never reach the state where real insight happens.
I wrote recently about the compulsion of unused credits, where I described how AI tools had turned me into a plate spinner, constantly offloading tasks and filling every freed-up moment with more work. What I didn't fully articulate there was what I was skipping in that loop. Each plate I spun carried an overhead tax: the mental cost of tracking it, checking back on it, deciding what to do with the output. Three agents running felt manageable. Eight felt like I was spending more energy managing the plates than doing any actual thinking. The gains aren't linear. Past a threshold, every new task you take on decreases your total output.
But the deeper cost was what I was skipping. I was skipping the strain. Every time I hit a hard moment and offloaded it to an agent, I was bypassing the exact cognitive process that produces my best thinking. The shower moment I described in that post, the idea that arrives uninvited during analog hours, that only happens when you've spent enough time under tension with a problem for your brain to keep processing it in the background.
Newport calls this the processor pipeline fallacy. Silicon Valley borrowed a metaphor from computer chip design: never let the pipeline sit empty, always keep something running. But a CPU can switch between unrelated commands in nanoseconds. A human brain needs 10 to 20 minutes. Treating yourself like a processor, always filling idle capacity, means you never reach the depth where the real breakthroughs live.
What gets lost
The Harvard Business Review coined the term "work slop" for this: AI-generated output that's produced quickly but is so shallow it makes everyone else's job harder. I think there's a personal version of this too. When I offload the hard thinking, the output I get back is competent but generic. It lacks the specific insight that comes from me having wrestled with the problem. The agent produces reasonable code, but it misses the architectural intuition that only surfaces after I've sat with the constraints long enough to feel them.
There's something Newport said that stuck with me: as AI handles more routine cognitive work, the ability to learn difficult things and sustain focus is becoming the scarcest professional skill. Not the most impressive, not the most rewarded (yet), but the scarcest. The people who can sit with a hard problem for an hour without reaching for the escape hatch will produce fundamentally different work than the people who can't.
Sitting with it
I don't have a productivity system for this. I'm not going to tell you to block your calendar or delete your AI subscription. What I've started doing is simpler: when I notice the urge to offload, I pause. I ask myself whether I'm reaching for the tool because it's genuinely the right approach, or because the blank page is uncomfortable and I want the discomfort to stop.
Sometimes the answer is that the tool is the right choice. Researching an unfamiliar API, generating boilerplate, exploring options I wouldn't have considered. These are legitimate uses that make me better.
But sometimes the answer is that I'm three minutes into a hard problem and I haven't given my brain the 20 minutes it needs to actually engage. In those moments, the most productive thing I can do is nothing. Sit with the discomfort. Let the gears catch. Trust that the strain is doing something, even when it doesn't feel like it.
The best ideas I've ever had didn't come from faster tools. They came from slower thinking. And I'm starting to suspect that the gap between the two is where the real work lives.