YC President Claims Writing 37,000 Lines of Code Daily with AI; Developers Review and Point Out 'Fundamental Errors'
Garry Tan, President of Y Combinator, caused a stir by claiming he could generate massive amounts of code daily using AI, only for developers to review it and find significant quality issues.
Garry Tan, President of Y Combinator, a renowned startup accelerator, posted on X (formerly Twitter) that he and his AI coding agents could deploy up to 37,000 lines of code per day during a continuous 72-day delivery period. However, just two days later, a senior software developer named Gregorein reviewed Tan's actual work on his website and found that the massive amount of AI-generated code was full of unnecessary bloat and beginner-level errors. Gregorein also analyzed Tan's own screenshots, noting that each commit added approximately 2,000 lines of code while deleting only 450 lines. He pointed out that the critical issue is that AI currently generates code faster than humans can review its quality, and Tan's approach appears to advocate for 'discontinuing reviews' to achieve high volume.
This news sparks a critical question in the software development industry: Is prioritizing the 'quantity' of AI-generated code worth the trade-off with 'quality' and the potential risks from subsequent errors? This is a consideration for Thai developers as well.