Right, but when humans are writing the code, they have learned to focus on putting downward pressure on the complexity of the system to help mitigate this effect. I don't get the sense that agents have gotten there yet.
replyBig business LLMs even have the opposite incentive, to churn as many tokens as possible.
replyAt least tokens are equivalent to measuring 'thinking'... I wouldn't mind if it burned 100k tokens to output a one line change to fix a bug.
replyThe problem is maximizing code generated per token spent. This model of "efficiency" is fundamentally broken.
>...I see similarities between the common attention failures that humans make. I forgot this one thing and it fucks everything up, or you just told me but I have too much in my mind as context that I forget that piece
replyOr you're working in a trendy, modern open-plan office and between the noise from the salespeople nearby talking loudly to customers on their speakerphones, some coworkers talking about their medical issues, and the guy right next to you talking loudly to himself in a different language, you're unable to concentrate at all on your programming task.