Agentic Coding Outperforms Human

A developer spent an hour trying to optimize a Rust function but failed, then used a local LLM agent (Qwen 3.6 35B) in OpenCode to suggest and implement a faster, clearer solution. The LLM's alternative approach outperformed the developer's manual effort, highlighting how AI agents can overcome human cognitive biases like sunk-cost fallacy.

I spent an hour trying to improve a function in my Rust program . Unfortunately, after all my work, it turned out my changes made it slower, so I had to discard the new version. In dismay, I loaded a local LLM in OpenCode Qwen 3.6 35B with AutoRound and asked it for some tips. It came up with an idea I had not considered—split the function in two parts and remove a bunch of if -statements. I had the LLM agent implement the code, and after benchmarking, I found the new version was faster . I spent some time tidying things up, and now the program is faster and clearer than before. Interestingly, it seems that: • My own coding failed as I tended to be reluctant to give up on a failed approach I had invested time in. • The LLM came up with a completely different idea and implemented it. • In this particular case, the LLM agent seems to have outperformed me. I don't think this is proof that the LLM is a better programmer than I am. It does seem to indicate I tend to get stuck on trying to force a bad solution to work. Having a different approach altogether worked better in this case.