There are more possible molecules than there are atoms in the Milky Way. Chemical space is unimaginably large. We will never make every molecule. We will never test every molecule. And this is where AI gets interesting. AI can explore chemical space at a scale humans never could. Generate structures. Predict properties. Rank candidates. Choose what to test next. But there’s an important distinction: Designing a molecule is not discovering a medicine, pesticide or material. Eventually, someone has to make it. Test whether it works. Then test whether it is safe. Toxicity. Metabolism. Exposure. Environmental fate. Effects on humans and other organisms. And those questions can’t be answered by a beautiful molecule on a computer screen. They require evidence. Experiments. Measurements. Data. And most candidates will fail somewhere along the way. That’s why I think the real AI revolution in science isn’t generating more possibilities. We already have 10⁶⁰ of those. It’s choosing better experiments from an almost infinite number of possibilities. AI accelerates the thinking. Automation accelerates the testing. Reality still gets the final vote.