Matthew Boston

Code was never the goal

January 19, 2026

Key takeaway: Measure your work by the problems you solve for users, and treat code as the medium you solve them in.

I love writing code, but code was never the goal. The point was always to solve real problems for real users. AI just sped that up.

It takes care of the boring parts, so engineers can spend more time designing systems, weighing tradeoffs, and making the decisions that change outcomes for users. If you define yourself by typing code, that can feel threatening. If you define yourself by the problems you solve, it’s mostly a relief.

Think about what makes a senior engineer valuable. It’s rarely typing speed or writing a sort from memory, which is why hiring on algorithm puzzles never told us much. It’s judgment. They know which problems to solve and which to ignore, when a simple solution is enough, and how to design systems that survive contact with reality.

None of that changes with AI. When generating code is cheap, deciding what code to generate becomes the bottleneck. A few things stay no matter who writes the code:

  • Systems thinking: how components interact, where bottlenecks show up, and how failures cascade.
  • Domain knowledge: your users, your business constraints, your regulatory environment. You can’t prompt that into existence.
  • Taste: picking the abstraction, knowing how much complexity is enough, and deciding where to trade speed for correctness.

The engineers who do well now will stop measuring themselves in lines of code and start measuring problems solved. They review the outcome, not the output. Code was always the medium, and now we have a faster way to work in it.