Future of Software is Growing the Code.

There will be endless data centers where code is no longer written, but grown. - Matrix, The Movie

"Humans won’t be writing code in the future." Surprisingly, most developers I’ve spoken to agree - it’s not a matter of if, but when. I’ve speculated on this future, and the most striking revelation? We won’t write code anymore; we’ll grow it. Let me explain, step by step.

In the future, humans won’t write code — they’ll define problems. They’ll craft specs, give examples, and guide machines. Their job? Expressing their intent.

AI can already write small programs in one go — input a problem, get a solution. But for anything more complex, it stumbles. Logic errors, syntax slips — it’s inevitable. The bigger the task, the more chances for mistakes. That's the nature of current LLMs.

The fix? Make AI work like developers — break problems down, generate incremental diffs, and iterate. Syntax error? Feed it back. Failing test? Do the same. Tools like "aider" and "gpt-engineer" already do this. More specialized tools are on the way, and OpenAI and GitHub are clearly steering us toward a no-code future.

New code creation is here: run 100 AI processes, generate 100 solutions, and benchmark them for speed, efficiency, and readability. Then, merge the best parts. I call it "growing the code." Want to keep all 100 versions? Fine — use them as starting points for future iterations. Evolutionary algorithms - I'm looking at you.

Code will explode—think Moore's Law for lines of code. Every million-line codebase could soon be AI-maintained, tested, and constantly evolving. Companies will soon compete on GPU/TPU hours spent optimizing their codebases.

It will happen one way or another, but what’s clear is that AI will take over the code-generation process, forcing us to adapt. As developers, we’ll move up the abstraction ladder, using AI to build and modify tens of thousands of lines of code daily — likely without even noticing.