CEOs Addiction to Vibecoding

For many CEOs, AI is a sophisticated form of procrastination. The dopamine hit of generating code replaces the harder, more valuable work of talking to customers and making decisions

For the past month, I’ve been losing two hours of sleep because of Codex and Claude Code. I feel like, if I don’t leave agents running in parallel, fixing things for my projects, I’m wasting the night.

Like most, I now have code to run my finances, organizes my calendar, remember personal details of the people I’ve met and check my company’s metrics. We all seem to build the same. I’m also pursuing personal projects I’ve always wanted to tackle. It’s fun. I’m learning.

It’s too much fun. It’s too addictive.

I spend hours implementing, refining, refactoring, polishing, tinkering. I ask the models questions, we build plans, and I witness their execution of big amazing things on my servers. It feels so rewarding.

I should be talking to users. My time is better spent thinking about my team.

“It’s teaching me about AI. The technical knowledge will make me a better CEO. I can understand Platzi’s dev team better now…” Maybe.

If I keep investing time into it, I’m sure I can achieve incredible results. But my job isn’t to become an expert in Codex, Claude Code, Cursor Compose, or whatever. I need to deeply understand what my customers want and need, understand what my team can build, motivate them around their talents, and connect all of that to an ambitious product vision.

My job is to talk to users, make difficult decisions, and build product.

AI is a miraculous casino. You don’t know whether the next result will be better or worse. You don’t know whether the time invested has a real ROI. You only know that if you put in tokens, something will happen.

My team is more talented at operating that slot machine, armed with technical context, human judgment, and experience. I can do it too, sure. But I should be appropriately inhabiting my own talent instead.

AI creates this feeling that we don’t need our teams all the time because, with a few words and enough tokens, we can feel progress.

Progress isn’t the same thing as building a better company.

I don’t regret the time I invested in learning AI. I’ll keep doing it. ButI don’t really believe in the idea of the vibecoding CEO and the one-human startup.

Companies are people, AI is a multiplier.

Some work will always be human: sales, leadership, vision. It’s on us to build an organization that executes, without the anxiety to push the next prompt.