The peril of laziness lost

Bryan Cantrill argues that LLMs fundamentally lack the programmer's virtue of 'laziness' — the drive to build crisp abstractions that simplify systems and reduce future cognitive load. Using Larry Wall's classic framing from Programming Perl, he contends that human time constraints force engineers to develop elegant abstractions, while LLMs, unconstrained by time or effort, tend to pile on complexity and bloat. He illustrates this with a critique of Garry Tan's boast of generating 37,000 lines of code per day via LLMs, pointing to an analysis revealing the resulting codebase was riddled with redundancy and junk. Cantrill concludes that LLMs are valuable tools but must be directed by engineers who still exercise the virtue of laziness — building simpler, more powerful systems rather than just more code.