Y Combinator CEO Garry Tan claimed he and AI agents produced 37,000 lines of code daily. A developer, Gregorein, examined Tan's AI-focused blog, revealing significant bloat and inefficiencies in the deployed code. Gregorein, an engineer with 13 years of experience, used an Anthropic Claude session to analyze the front-end code.

The website, purportedly built with AI, makes 169 server requests totaling 6.42 megabytes for assets when a user visits. This contrasts sharply with the Hacker News homepage, which requires seven requests for 12 kilobytes. The site ships 28 test files (300 kilobytes) to every user and loads 78 JavaScript controllers for features not present on the homepage, all downloaded 'just in case.'

Further issues include the logo being downloaded in eight formats, including an empty file. Large, uncompressed PNGs are used instead of modern, smaller formats, wasting approximately 4 megabytes. Gregorein also identified duplicate content, an empty CSS file, bloat from a rich-text editor on a read-only page, missing image descriptions, and analytics code designed to bypass ad blockers. This analysis suggests that while AI accelerates code generation, code quality and review remain paramount, as unchecked output can lead to performance issues and fragility.