Coding is dying, and so are codebases

Code is dying, not just because the act of coding is being eaten by AI, but because codebases will die as they will no longer be a primary artifact.

First principles time!

Where do apps come from? A builder's initial vision, and the accumulated feedback of users reacting to it. That's it.... the codebase (including tests) is a "constraint solving phenotype" of those concerns, not something 'primary' in and of itself.

The codebase ascended in importance because of the time and capital cost to produce it. When strong coders require strong capital to hire, and weeks upon weeks to work, you treat each line of code as a sacred form of time-and-capital caching.

Today, we have PhD++ level AI engineers, always on and increasingly too cheap to meter... unconstrained by time and money, you could simply re-write the product instantaneously in response to any new feedback.

Entire apps will flow from an LLM responding to aggregate feedback, coughing up something amazing and complete from its latent space. This ultimately relies on big contexts, strong foundational models, and reasoning either baked-in or bootstrapped agentically, but the trends are clear.

Code is best thought of as cached reasoning. But are you caching for lead-time and capital constraints, or caching for usage latency? We are moving from the former to the latter, and with that transition the idea of the codebase itself dies.