Jagged Activation Energy

The term "activation energy" comes from chemistry, where it means the minimum amount of energy required for a chemical reaction to occur. A catalyst is simply something that lowers the activation energy, and catalysis is the ensuing increase in reaction rate caused by the lowered energy requirements. This is a handy metaphor in tech and popular for good reason.

Software development is obviously deep in the throes of catalysis right now. Everything is changing really fast, and everyone is racing to predict what will happen. The cost of writing code has fallen through the floor. So are engineers the 21st century stevedores or will we be the latest case study on the Jevons paradox? There is an abundance of pieces you can read on that topic, and this isn't one of them.

The point I want to make: catalysis is an even better metaphor than most people realize. A catalyst doesn't lower activation energy uniformly, but rather in one particular step in the reaction. The bottleneck moves to the next-slowest step. The reaction is as fast as the slowest step in the process.

When you make something cheaper, people use more of it. So where's all the software? The answer is obvious: writing code is not the only step in the chemical reaction of software development. The frontier of activation energy is jagged in the same way that the frontier of AI capabilities is. Coding might be profoundly cheaper, but deployment and distribution haven't moved much.

This creates cliffs: now-cheap steps that sit right next to still-expensive ones. Those cliffs are where the opportunities are. When coding was the bottleneck, there wasn't a lot of incremental value in making the other steps cheaper. Now that coding isn't the bottleneck, it's hard to improve software development holistically by vibe coding even harder. You are just catalyzing a reaction whose rate-limiting step is now downstream.

The biggest gap right now is between creating something and getting it in people's hands. Most of what we have today was built with the underlying assumption that coding is expensive. It's not true anymore.