Hacker News front pageJun 11, 2026, 20:26 UTC+8

Lines of Code Got a Better Publicist

David Curlewis argues that Google, Anthropic, and OpenAI are all touting volume metrics like 'percent of code written by AI,' which is just lines-of-code counting with better PR. He contrasts earlier outcome claims (Copilot made tasks 55% faster) with today's unfalsifiable adoption numbers that rise regardless of real improvement. The post walks through conflicting research: METR first found experienced devs 19% slower with AI, then walked it back and abandoned the study design; an NBER survey of ~6,000 execs found ~90% reporting no measurable productivity impact. Anthropic simultaneously claims '8x more code' and published an RCT showing 17% lower comprehension with no significant productivity gain. Curlewis worries these numbers are driving layoffs—Block cut 40% of staff, Atlassian cut 10%, both explicitly citing AI as the rationale.

Why it matters: A sharp commentary with concrete industry numbers, reframing 'AI wrote X% of code' as repackaged lines-of-code metrics. Hits all three HKR axes. Not scored higher because it's an opinion piece rather than a primary release, but the take is pointed and substantive enough for fe...

Read the original ↗Export Markdown

Take

This post lands because it strips the paint off a claim you've heard everywhere: Google, Anthropic, and OpenAI all say AI writes 75-80% of their code. David Curlewis points out the obvious—this is the same lines-of-code metric we spent two decades mocking, just with a better publicist. His contrast is sharp: a few years ago GitHub claimed Copilot made tasks 55% faster, a falsifiable outcome claim. Today's percentage numbers can't fail; they rise regardless of whether anything actually improved, because adoption is the one thing everyone agrees is real.

The research conflicts he walks through are the useful bit. METR first found experienced devs 19% slower with AI in their own codebases, then walked it back in February 2026—developers now refuse to work without AI, so clean measurement is impossible. An NBER survey of ~6,000 execs found roughly 90% reporting no measurable productivity impact. Anthropic simultaneously claims engineers ship '8x more code' while their own RCT showed 17% lower comprehension and no significant productivity gain.

I'd discount this a bit: it's a personal blog, not a systematic review, and the studies span different methods and populations. But it nails why 'percent AI-written code' is a vanity metric—and Block cut 40% of staff, Atlassian cut 10%, both explicitly citing AI, with numbers like these as the rationale.

Source: Hacker News front page · curlewis.co.nz
Published: Jun 11, 2026, 20:26 UTC+8 · Jun 11, 2026