What we believe
The code is not the work. The code is the crystallization of understanding. Production capacity has lapped understanding capacity, and only two moves are available: lower what we mean by software, or move the primary artifact out of the code and into the understanding itself.
This work is the second move.
For developers
AI can be the amrita or the poison. Used to produce more commodity code, faster, it accelerates decay. Used well, it gives back the part of your head that used to hold syntax and framework trivia, and that space is worth spending on what was always the hard part: what the business actually needs, how the system actually behaves, and whether the tests describe the requirements or just the implementation.
That takes discernment, more of it than the job has ever asked for.
For training
Teaching syntax is close to worthless now. Teaching judgment isn’t. Our courses still go deep on HTTP, messaging, testing, state, and delivery, because you can’t judge code you don’t understand. What’s changed is how they’re taught. The instructor writes the competing approaches by hand, including the ones we don’t take, so you can see why one fits. Once the pattern is clear, the AI does the typing. And the rule throughout: never accept a suggestion you don’t understand.
More in how the courses are taught, and in the working list of ways to use AI to learn instead of just to produce.