7 ms·

and it's wrong. a 6000 line class is not easier for a model to understand. the same things that help humans also help agents. I find myself adding linters tha

and it's wrong. a 6000 line class is not easier for a model to understand. the same things that help humans also help agents. I find myself adding linters that must pass and the agent muss fix that limit file size, function length, function complexity, how many files in a directory. a little more work for the agent, but the codebase is healthier and the agents write fewer bugs.

lukevdp 4mo ago
I don't think the same things that help humans help agents. Simplicity helps humans, for agents parsing complexity is a breeze. Not saying code quality isn't important - it is. But I think what is described as quality code will change.
- cjbgkagh 4mo ago
  Agents still pay a penalty for complexity even if it is a smaller one.
  - digitaltrees 4mo ago
    Parsing single file is easier than navigating a file system for an LLM. Until the models have context windows large enough to hold the entire codebase in one shot, single files will beat multiple files every time.
    - bossyTeacher 4mo ago
      This. I suspect the codebases in the future will be made of a small number of gigantic source files. These will be able to be transpiled into a more human friendly that produces multiple smaller files per big file in human-debug mode.
      - cjbgkagh 4mo ago
        As a human who typically uses large files, 10k to 30K lines of code files are pretty common, I find the agents don’t read the whole file after the first time, they almost always do a range select for the bit they are interested in.
        digitaltrees 4mo ago
        But you wouldn’t argue that a 30k line file is good code would you? Humans write bad code to. For me the litmus test is: will someone read this in the future. If yes, then write good code. I don’t think we are in the era when a human will never read the code again in human history. So we should still strive for human intelligibility.
        cjbgkagh 4mo ago
        The code I write is good. I don’t write for the lowest common denominator, readers need to be hard core functional programmers as a pre-req. The great thing about AI is that it’s raising the bar on lowest common denominator.

lukevdp 4mo ago

I don't think the same things that help humans help agents. Simplicity helps humans, for agents parsing complexity is a breeze. Not saying code quality isn't important - it is. But I think what is described as quality code will change.

cjbgkagh 4mo ago
Agents still pay a penalty for complexity even if it is a smaller one.
- digitaltrees 4mo ago
  Parsing single file is easier than navigating a file system for an LLM. Until the models have context windows large enough to hold the entire codebase in one shot, single files will beat multiple files every time.
  - bossyTeacher 4mo ago
    This. I suspect the codebases in the future will be made of a small number of gigantic source files. These will be able to be transpiled into a more human friendly that produces multiple smaller files per big file in human-debug mode.
    - cjbgkagh 4mo ago
      As a human who typically uses large files, 10k to 30K lines of code files are pretty common, I find the agents don’t read the whole file after the first time, they almost always do a range select for the bit they are interested in.
      - digitaltrees 4mo ago
        But you wouldn’t argue that a 30k line file is good code would you? Humans write bad code to. For me the litmus test is: will someone read this in the future. If yes, then write good code. I don’t think we are in the era when a human will never read the code again in human history. So we should still strive for human intelligibility.
        cjbgkagh 4mo ago
        The code I write is good. I don’t write for the lowest common denominator, readers need to be hard core functional programmers as a pre-req. The great thing about AI is that it’s raising the bar on lowest common denominator.