The agent could not fully understand lukecolman.com because the site uses client-side rendering with JavaScript; it retrieved only HTML scaffolding with no visible content text. The agent assembled a partial answer from prior knowledge (that Luke Colman is a frontend developer), image asset names inferred from the HTML (enviopack, ultimate-comics), and social media links found via web search. The site is poorly agent-readable due to its heavy reliance on client-side rendering, leaving the agent unable to extract core content like project descriptions, pricing, or service offerings directly from the domain.

• ›Steps [9] and [11] were the only content-bearing steps: [9] fetched the Netlify mirror (same HTML scaffold as the .com version) and [11] found social media links and confirmed the developer's identity via web search, but neither returned actual page content or project details.
• ›The agent inferred what the site *is* (a Next.js portfolio for a frontend developer) from preloader animations and image asset filenames in the raw HTML, not from readable prose or structured metadata. It had to rely on prior knowledge and web search to establish that Luke Colman is Argentine and works as a frontend developer.
• ›The site publishes no machine-readable data about what it offers, who it's for, or how it's priced—because it is a personal portfolio (not a commercial product), but the agent could not discover even the portfolio's project list, descriptions, or contact methods. Both direct navigation attempts ([6], [7]) returned 404s, indicating the site is a single-page app with no crawlable structure.
• ›The agent explicitly noted it could not see 'full content visibility' or 'specific project details' and flagged that visiting the live site with JavaScript enabled would be required. This is a fundamental gap in agent-readiness: the site does not serve its core information to non-JavaScript clients.