There has never been a better time to learn to code. There has never been a worse time to learn to code. The gulf has never been wider. The barbell has never been more stretched. The middle is squeezed yet again.

At one end of the barbell, a new class of high-leverage engineer will emerge—those who can whisper the right words to models and push them to a place in latent space where the tokens flow. In the middle, the average who lack agency to learn or the stubborn who lack openness to change will be pushed between a rock and a hard place. On the bar end, AI agents will be synonymous with the generation of common patterns that will be never typed again by human fingers, or in the worst case, only via tab complete. The distribution is not normal. As AI scales our intelligence and speed, the distribution is power law—the top 10% will capture 90% of the value.

Learning how to code is no longer typing syntax. It is something else entirely. It is about understanding logic and patterns, about communicating clearly to instrument AI to improve upon itself. It is about reviewing pull requests and distilling the gold from the slop. There has never been a better time to code, for we can all be engineers, though the definition of an engineer is scattered into smithereens, emerging as something entirely new.

We have been given an infinitely patient tutor at our fingertips. It is about asking the right questions and iterating on them. We are ushered into an era where anyone can create, and “developer” is no longer a coveted title held only by university graduates. With the right aesthetics and judgement, you can will your ideas into the world via a compelling sequence of natural language. This is the new grammar of creation.

Coding now is about picking the right models to do fitting work and stringing together APIs. The question isn’t whether you should learn to code—it’s whether you can afford not to understand the new paradigm. The middle is thinning and you will be pushed. Choose a side or have a side chosen for you.