Guillermo Rauch raises important questions about the current state of artificial intelligence in software engineering, especially in light of recent advancements where AI can write entire compilers and web browsers. He challenges the notion of whether these developments genuinely represent progress or if they're merely interesting stories. With over 25 years in programming and a decade leading Vercel, Rauch draws on his extensive experience to argue that real achievements in AI should tackle complex engineering problems that have eluded humans or rarely been solved. He outlines several ambitious tasks that would signal true AI advancement. One example is identifying vulnerabilities similar to the Heartbleed bug and creating proofs of concept. Such an achievement would not only enhance cybersecurity but also demonstrate AI's capability in both offensive and defensive roles. Another task involves rewriting the TypeScript compiler in Rust with better performance. This would show that AI can handle intricate software systems, an area where even skilled human engineers have struggled. Rauch also discusses the challenge of accurately re-implementing Apple's Liquid Glass design system using AI, which would indicate a sophisticated understanding of visual and functional elements in UI engineering. Identifying and fixing data integrity violations in major open-source databases, like those exposed in Kyle Kingsbury's Jepsen audits, is another critical task. Successfully addressing these issues could prevent significant data loss or privacy breaches. Finally, he addresses the potential for AI to develop new image compression formats that surpass current technologies like WebP and AVIF, which could significantly reduce bandwidth usage across the internet. Each of these tasks, if achieved by AI, would mark a significant leap towards true super-intelligence in software engineering.