Karan Notes, half-formed theories, and small things I wanted to remember without making a whole performance out of them.
In July I guessed my agent waits were about forty seconds, and built a whole practice around making forty seconds pleasant. This week I measured it. I was right, which turned out to be the problem: the number wrecking me was the two hundred times a day it taps me on the shoulder.
I got bored waiting for an agent and asked what Rick Rubin would do with all this. He can't play an instrument. I can't code. The job we're both left with is picking, and picking doesn't respond to effort.
I'd been sending my wife screenshots all week, mostly of things that were broken. At dinner she said it's sad how much I want to automate, and that it could work. The second half is the interesting bit.
Last night I changed my Instagram to a black circle and wrote 'unsubscribed from society' in the bio. With a wink. Then I couldn't be bothered to do it anywhere else. The wink was the tell.
An AI asked what percentage of my time I'm budgeting towards being free, alive and joyful versus making money. I did not have a number. Which is when I noticed the accountant in my head has never once shown me his working.
Listening to DeMello, a line landed sideways: cling too hard to being a genius and you're the actual fool. I tried swapping the label for madman. It was slightly liberating. Only slightly, which was the honest part.
I built the digital army I'd been describing to myself for months, then sat down at the desk and got bored anyway. Which forced a stupid question: why do bosses exist at all?
The agent is off doing the thing, and I'm sitting there like a dog that's been told to stay. On discovering the bug was never distraction, it was boredom, and the fix was never focus, it was play.
An agent did four hours of my work in six minutes, and I spent those six minutes checking three apps in rotation. On losing the grinding kind of focus and noticing the world stopped asking for it.
I have been building software nobody else will ever use, and ignoring the little voice that asks if it could be a product. The ignoring turns out to be the interesting part.
A hard week, one good half of football, and a reminder that maybe the whole point is to take the brake off and play your way.
I was building an early customer validation system when I realized I did not really care about the customer I was meant to understand. Maybe the answer is not artist over market, or market over artist, but choosing a customer whose world you actually want to enter.
I tried to brainwash myself out of the billionaire chase. Not in a cult way. More like building a tiny belief laundrette for the parts of my mind still waiting for permission to relax.
I asked AI to teach me Jung better than most psychologists, then noticed the old desire to be exceptional hiding inside a simple afternoon of curiosity.
I made writing for this site so easy it started to feel suspicious. But maybe the cheating is not in using AI. Maybe the cheating is removing the friction between noticing something and leaving a trace of it.
On holiday, I tried tearing down every authority figure I had absorbed and ended up with one useful thought: everyone is a monkey playing games.
I refused to get into a temple purification pond full of fish, then immediately used AI and philosophy to make myself feel better about it.
On holiday, I kept thinking about the learning app and realized the plan might be annoyingly simple: serve 20 to 50 people well, prove the loop, then scale what works.
At lunch, I noticed a small status game still running in me: the urge to prove intelligence, taste, depth, and maybe even how little I care about status.
AI is making copying faster, cheaper, and more inevitable. So maybe the moat is not just execution. Maybe it is attention, taste, and a company having a flavour that can only come from the authentic nature of its leadership.
I thought I was learning AI agents. But after a week of getting my hands dirty, the real thing I noticed was stranger: maybe the business bottleneck is no longer permission, capital, or even me.
Watching Y Combinator talk about “token maxing” made me realize I was still using AI with a scarcity mindset. If intelligence is becoming abundant, the real bottlenecks are no longer tokens or effort, but aim, taste, review, and iteration.
The interface for work is shifting from clicking around apps to expressing clear intent into connected systems. If that is true, productivity becomes less about operating software and more about thinking clearly enough to direct it.
Learning starts to change when you stop watching from the stands and get back on the court.