Hacker News Reader

I have found that claude (opus 4.8 and 5) can do a very good job of implementing good architecture and system design if I can just steer it right. left to its own devices it does produce less than optimal code, but if I just spend a few minutes explaining how some pattern will improve the code, claude will go do it the right way, and can even have a pretty sensible discussion about the tradeoffs involved. so it does know about good code architecture and patterns, it just doesn't have the taste and initiative to use them without being told.

> left to its own devices it does produce less than optimal code, but if I just spend a few minutes explaining how some pattern will improve the code,

But that requires actually reading the output, which I am pretty certain only a rounding error of programmers are doing at this point.

And the next time your agent comes across the same problem, it will likely use the same previous inefficient way again (despite all attempts to codify the instruction in AGENTS.md, CLAUDE.md, skills, whatever)

> they are very strongly tuned to value code that has no errors which can be picked up by linters or similar, while humans work the opposite way

This is a really valuable insight! It resonates well with my own experience revewing AI-authored code: I look hard at interfaces, architecture and performance, and merely glance over code that just "gets the job done", because if it works, I'm not worried about it. Now I can express that dichotomy much more clearly -- thank you!

> Where LLMs excel is in code-level bugs (as opposed to system bugs, design bugs, architecture bugs, integration bugs, etc).

Blame the benchmarks game. They're optimizing for that and that's what those things are measuring.