June 11, 2026
More code, less clue
Lines of Code Got a Better Publicist
Tech bosses are bragging about code volume, and the comments are absolutely roasting them
TLDR: AI companies are boasting that most new software is now written by AI, but critics say that’s just a flashy volume stat, not proof anything improved. In the comments, readers mocked the obsession with giant code counts and asked the obvious question: does any of this actually help users or make businesses better?
The big fight here is deliciously simple: are companies measuring real progress, or just counting words on a screen? The article argues that giant AI firms like Google, OpenAI, Anthropic, and Cursor are flexing stats like “75% of code is AI-written” and “100M lines a day” as if bigger automatically means better. But plenty of readers were not buying the hype. Their main complaint? This sounds like the software version of judging a chef by how many onions they chopped, not whether dinner tasted good.
That mood came through loud in the comments. One reader summed up the whole mess with a mini three-act tragedy: before AI, the goal was make more money; with AI, the goal somehow became ship more code; and after AI, everyone will sheepishly remember the goal was making money all along. Another commenter mocked a survey where 219 managers gave 219 different definitions of “AI-native engineering,” basically saying: give enough people a text box and of course you get chaos. And then came the real eye-roll fuel: outrage over a flashy OpenAI post about a product built “100% by agents” that supposedly never clearly explained what the product even does.
The funniest running joke in the thread was the fear that executives have fallen in love with absurdly huge code counts, including the now-infamous dream of 1 million lines of code per engineer per month. To many readers, that doesn’t sound like innovation. It sounds like satire accidentally promoted to corporate strategy.
Key Points
- •The article says AI coding vendors increasingly promote volume metrics such as percentage of code generated by AI or lines of code written per day.
- •It contrasts those claims with earlier outcome-based messaging, such as GitHub’s claim that Copilot helped developers complete tasks 55% faster.
- •The article cites mixed research findings, including positive results from Cui et al., quality concerns from GitClear, and changing conclusions from METR.
- •According to the article, METR later said AI probably speeds developers up in 2026 but that precise measurement has become harder because developers now rely on AI and agentic work is difficult to time-report.
- •The article cites an NBER survey reporting broad AI adoption among firms but limited measurable productivity impact, with cross-study organizational gains around 10%.