Your code is worthless

Forty years after Apple programmer Bill Atkinson wrote "-2,000" in his weekly lines of code report after elegantly shrinking bloated software, Silicon Valley has circled back to the same flawed metric. Y Combinator's CEO recently boasted about AI agents cranking out 37,000 lines of code daily as proof of productivity, but this industrial-scale code generation is creating what one critic calls "technical debt financing" rather than genuine value.

The fundamental problem remains unchanged: source code itself holds zero intrinsic value. Microsoft could open-source Windows tomorrow, but users wouldn't flock to a startup selling identical code because they're paying for outcomes, not line counts.

AI-generated code mountains actually slow human developers by clogging feedback loops with dependencies and noise, turning what should be quick iterations into 20-minute build cycles that shatter focus. When interruptions force developers out of flow states, it takes an average of 23 minutes to regain focused productivity.

The real winners in this "agentic era" won't be companies burning the most AI tokens, but those measuring what actually matters: whether users' lives improve, or whether repositories just got bigger.