The Day AI Actually Scared Me

I've spent decades accumulating skills, pattern recognition, relationships, and the judgment that comes from making decisions when the answer isn't obvious.

For a long time, AI felt like a multiplier for those capabilities. Chat was already making me substantially faster at thinking through problems, drafting work, and testing ideas. I thought I understood the direction of travel.

Then I watched the pace of AI-assisted software creation accelerate. Claude Code and the broader movement toward agentic development were important parts of that shift for me. Later, the story around Claude Cowork sharpened the realization: tools were no longer just helping someone complete a task. Increasingly, AI was helping create and improve the tools themselves.

The precise timeline of any one product matters less to this story than the feeling I had watching the cycle compress. Work I associated with weeks of coordinated engineering effort could sometimes be explored in hours. I was not looking at a better autocomplete. I was looking at the early shape of a different operating model.

It scared me.

Not because I concluded human expertise was worthless. Quite the opposite. I realized how much of my professional identity had become attached to the difficulty of producing something, rather than the quality of deciding what should be produced.

There were two choices. I could dismiss what I was seeing as a novelty, or I could rebuild my own habits around it.

I chose the second. Hundreds of hours became thousands of hours of experimentation: building, breaking, revising, learning how to describe problems precisely, and learning when not to trust an answer. The hardest part wasn't memorizing prompts. It was letting go of workflows I had spent years mastering.

There is an uncomfortable stage in becoming AI fluent. You know enough to see the potential, but not enough to operate confidently. You oscillate between amazement and frustration. A remarkable result one hour can be followed by a ridiculous failure the next.

That tension is healthy. It forces a new discipline: verify outputs, define constraints, preserve context, and remain accountable for the result. AI can accelerate execution, but it does not absolve anyone of judgment.

I still don't consider myself AI native. I am AI fluent, and working to become more so. My children may grow up with a relationship to AI that feels as natural as the internet eventually felt to my generation. I remember waiting for a modem to connect. They may never understand why we separated 'using software' from 'asking software to do something.'

The fear that changed me was not that AI would make me irrelevant. It was that standing still would.

That is the lesson I keep returning to: the advantage isn't knowing everything. It is being willing to relearn how you work before circumstances force you to.