Who's in charge, me or Claude?

I have a paid Claude account. Before that I paid for OpenAI and Gemini too. Lately I've noticed a strange thought creeping in at the end of the day: can I afford to stop? Somewhere, an extremely capable model is sitting idle, on a plan I'm paying for, with usage that resets on a timer whether I use it or not. Every hour I don't give it a task feels like money left on the table. So should I be queuing research before bed? Should I wake up every sixty minutes to check whether it has room for another job? And if I do, who exactly is working for who?

The meter is always running

I spent years in DeFi, where idle capital was the enemy. Liquidity sitting in a pool and not being traded against was a cost, and the whole industry obsessed over capital efficiency. Every unit should be working. That mindset is useful when you're designing a protocol. It turns out to be corrosive when you apply it to your own life.

Because that's what the subscription quietly does. It turns the AI into idle capital, and it turns me into the person responsible for deploying it. The tool doesn't ask for anything. It just waits. And the waiting is the pressure. A gym membership makes you feel guilty for not showing up. This is the same guilt, except the gym can also do your research, write your code and draft your strategy, and it's never tired.

When the tool starts setting the schedule

Here's where it flips. The moment I start waking up to feed it, the hierarchy has inverted. I'm no longer deciding what's worth doing and then using the AI to do it. I'm looking for things to do so the AI has something to work on. That's busywork, just at a higher price.

And it doesn't even work. Every task I hand over comes back as something I have to read, judge, correct and decide on. The AI's capacity is close to unlimited. Mine isn't. Push more work through it and I become the bottleneck, reviewing output at two in the morning that I wouldn't have commissioned if I'd been thinking clearly. At that point it's not my assistant. It's my manager, and it sets a relentless pace.

What's actually scarce

The mistake is measuring the wrong resource. AI capacity is abundant, and it gets cheaper every few months. What's scarce is my attention, my judgment, and the quality of the decisions I make about what to point all that capacity at. Idle AI is cheap. A tired person making decisions is expensive.

The best questions I've asked an AI didn't come from staring at it. They came on a walk, in the shower, in a conversation about something else entirely. Rest isn't time stolen from the machine. It's where the direction comes from. Without it, I can keep the model busy around the clock and still go nowhere, very efficiently.

So I've started treating it differently. I give it the long, well-defined jobs before I step away, not instead of stepping away. I judge a day by the decisions I made, not the tokens I used. And I let the capacity sit unused without feeling like I lost something. A car parked in the driveway isn't wasted. It's ready.

Bottom lines

A subscription turns AI into idle capital. That framing quietly makes you responsible for keeping it busy.

If you're finding work for the AI, the hierarchy has flipped. Decide what's worth doing first, then use it.

Your review is the real bottleneck. More tasks in means more judgment out, and judgment doesn't scale.

Measure decisions, not tokens. Unused capacity costs little. A bad call made tired costs a lot.

So who's in charge? The honest answer is whoever decides what's worth doing. The AI will happily do anything, at any hour, and it will never tell me to stop. That's exactly why the decision to stop has to be mine. The question for this moment isn't how to keep the smartest tool in the world busy. It's how to stay the one who decides what it's for.