20 Years Building the Web, Now Training Its Replacement

The author's 20-year narrative traces a fundamental shift in how web development knowledge is produced and consumed. In the mid-2000s, the mission was explicit: improve the web for users and developers through adherence to standards, elegant code, and open-source sharing. That collaborative spirit—thousands of deployments, millions of lines of shared code, extensive technical writing—created a vast commons of knowledge. The irony the author highlights is that this knowledge commons became the training ground for the very AI systems now disrupting the profession. LLMs are a direct product of developer labor, documentation, and open-source contributions.

Yet the author resists a narrative of simple obsolescence. Instead, he suggests a distinction between how AI solves problems (through computational brute force and pattern matching) and how humans solve them (through design principles, elegance, and simplicity). The example of aligning Figma and Webflow naming conventions—where discipline and consistency eliminate the need for manual measurement and error—illustrates this difference. It is not a claim that humans will always outcompete AI, but rather that certain kinds of problem-solving (those that prize coherence, restraint, and principle-driven design) may retain value precisely because they resist automation through scale.