Twenty hours across three days

Since gaining access to Gemini 3 Pro the day before yesterday, I have spent far more time using AI. It is not just the model’s capabilities; I no longer feel I must hold back every idea.

I told AI I wanted to build a physical robot dog with Gemini 3 Pro as its brain. It suggested the project might be feasible and showed me a preview of how it could look and interact. I could ask for changes without worrying about testing someone’s patience. Yesterday I designed a prototype of my “CyberPup,” and in a day and a half I finished the first stage of the code.

Today, while exploring GitHub, I discovered that I could set up OpenClaw on my own computer more easily than I had expected. I spent nearly the whole afternoon interacting with it. I even made an AI helper that could post to a group chat and collect data from web pages. I feel I have only begun to explore what I can make with these tools.

Before using AI much, I could not understand why a chip company such as Nvidia was valued so highly. Around the start of 2023, I suspected a bubble, perhaps partly because I had missed the gains. Over the years I had also been struck by notable AI applications. As a user I began to question my earlier judgment, try AI products, and buy some AI-related stocks.

Paying for a month of Gemini 3 Pro sharpened that feeling. I still think there may be an AI bubble, but I no longer accept the most sweeping version of that claim. In my experience, these tools have improved what I can produce over the past three years by more than I once imagined. I now see making things—and paying for useful tools—as important ways to learn what is really possible.