The agent assembled a reasonably complete answer about Katapult Labs' positioning, service offerings, and target market by relying heavily on prior knowledge and web search results rather than the actual website content. The site itself proved largely inaccessible—the homepage is client-side rendered (Next.js) and returned only boilerplate HTML headers, forcing the agent to reconstruct the narrative from job postings, search snippets, and domain knowledge. Despite this friction, the agent delivered a coherent explanation of what they do, who they're for, and how they differ from alternatives, though specific pricing and service details remain vague.

• ›Steps [1] and [3] fetched the homepage but returned only HTML scaffolding—no actual content was parseable. The agent could not extract details directly from the site itself.
• ›Step [12] (contact page) similarly returned only HTML structure with no meaningful content about services or pricing.
• ›The agent relied on 50% prior knowledge and 33% web search (steps [8], [9], [14], [15], [17], [18], [20]) to fill gaps. Search snippets and job postings became the primary source of ground truth—these were not fetched as full pages but used as references in the final source list.
• ›Pricing information could not be found anywhere—the agent correctly identified this as a confusing gap and noted that Katapult requires custom consultation.
• ›The service framework (0→1, Team Augmentation, Agentic Transformation, Ventures) appears to come from prior knowledge or search snippet inference rather than directly from website prose.
• ›The site's Next.js architecture made it opaque to content extraction; the agent had to work around this limitation by pivoting to external sources.