I got tired of fixing broken CSS selectors, so I bypassed the DOM entirely using AI

If you've ever built a web scraper, you know the honeymoon phase doesn't last long. Writing the initial script with Beautiful Soup, Cheerio, or Puppeteer is fun. But then, a few weeks later, the target website pushes a minor UI update. Suddenly, your script breaks because they randomized their Tailwind utility classes, nested a <div> one level deeper, or changed a .price-tag to .price-container. You open your IDE, inspect the new DOM, update your XPath or CSS selectors, and push the fix. Rinse and repeat. Web scraping isn't hard. Maintaining scrapers is a nightmare. The "Aha!" Moment I was managing a pipeline that scraped e-commerce data and directories. I realized I was spending 80% of my time maintaining brittle selectors rather than building new features. I asked myself: Why are we still traversing the DOM in 2026 when LLMs can understand the context of a page? What if, instead of telling the script where to look (via XPath), we just tell it what we want? Building a Selector-F