Its hilarious to me that before AI, I would suggest we refactor our code base to make it more readable and easier to understand and would get all the excuses. “Oh we don’t have the time”. " we could introduce bugs". “We would need to test that”
Now when I mention to make the code easier for an LLM to understand, they’re chomping at the bit for it.
Still you can see why right? Improving docs/understandability for AI is good for them - they get more capable AI. Improving it for humans only helps other people, which is less of a priority.
I also got to refractor a major headache away, it was finally cheap enough to do all in one go instead of putting in a new pattern and incrementally change. The “Do one, do one, now repeat” prompt pattern works really well for these.
Its hilarious to me that before AI, I would suggest we refactor our code base to make it more readable and easier to understand and would get all the excuses. “Oh we don’t have the time”. " we could introduce bugs". “We would need to test that”
Now when I mention to make the code easier for an LLM to understand, they’re chomping at the bit for it.
Yeah… when the only documentation is CLAUDE.md…
Still you can see why right? Improving docs/understandability for AI is good for them - they get more capable AI. Improving it for humans only helps other people, which is less of a priority.
Before AI, refactoring was expensive, high risk, unclear payoff.
With AI, it’s cheap, higher risk, obvious payoff.
The obvious payoff being that engineers with an actual work ethic get to fix the mess.
I don’t care about work ethic, I want clean and tested code, that performs well and is easy to extend and replace
You get none of the above without someone willing to do the work.
I also got to refractor a major headache away, it was finally cheap enough to do all in one go instead of putting in a new pattern and incrementally change. The “Do one, do one, now repeat” prompt pattern works really well for these.
cheap at first