The Case for Ambitious Software Architecture

The current software landscape is increasingly saturated with low-quality, AI-generated projects—described by Dan Luu as "vibeslop"—that often rely on misleading benchmarks. However, as Luu first reported via Hacker News, builders now have the capability to be more ambitious with the software they create, moving toward high-quality systems that were previously deemed too large to be feasible.

One example is pgrust, developed by Michael Malis. While many rewrite projects are dismissed as low-quality, Luu notes that Malis has employed rigorous testing with pgrust that has uncovered actual bugs within Postgres, including memory safety issues in the dictionary loader, a money overflow bug, and numerical bugs affecting SP-GiST operations. Luu suggests that even if pgrust as a product is not successful, the pursuit of its performance optimizations could durably improve Postgres and Clickhouse.

Luu further illustrates this shift in capability through personal projects completed in a fraction of the time previously required. Using SOTA models, Luu developed a regex engine, a faster ripgrep fork, and a version of Rust with improved build speeds in a low single-digit number of hours. Additionally, Luu spent approximately 20 hours developing Azul AI, a superhuman matched time control agent, to make it the strongest AI at the time, before Gabriel Cangussu developed a superior AI in even less time using Fable 5.1 and Opus 5.5.

While reaching a professional open-source quality bar still requires significant effort, Luu highlights that the speed of development has accelerated to a point where developers can architect for higher performance and greater ambition.