The agent assembled a coherent explanation of Badger's business model by combining fragments from the official website and external sources, but encountered significant friction: the site itself was minimally informative (Wix-hosted pages with truncated HTML responses), key information like pricing was completely absent, and the FAQ page returned 404. The agent had to rely heavily on web search results and third-party coverage to construct the answer, making the site itself poorly agent-navigable despite the run succeeding.

• ›Steps [4], [11], and [17] were cited as sources, but none returned substantive content—they were Wix-hosted pages with truncated HTML. The agent's understanding came primarily from search result snippets and external sources (EPFL startup profile, Watson news article, app store listings), not from the domain itself.
• ›The site does not publish pricing, feature comparisons, or 'how it works' mechanics in a machine-readable or even human-readable form. The agent explicitly noted: 'specific pricing is not publicly available on their website' and 'The FAQ link from search results is broken.' This forced reliance on inference and external reporting.
• ›The agent had to assemble meaning from fragments: search results pointed to app store pages, a news article, and an EPFL PDF; the main site was functionally opaque. 67% of fetches were sourced from web search, indicating the domain itself is not self-documenting. The site's positioning, differentiation, and business model are not clearly articulated on badgercards.ch.
• ›The agent correctly identified a strategic tension (dedicated app vs. wallet integration) by comparing Badger to industry alternatives sourced externally, not from the site itself. The site provides no competitive positioning or explanation of why its approach differs.