The agent assembled a partial but substantive explanation of Bodhi by working around the target domain's complete lack of public documentation. The bodhi.wtf site itself is a bare React shell with no visible content across all routes, forcing the agent to rely entirely on external sources: a blog post explaining the protocol and philosophy, a RetroPGF Hub project entry, and the X/Twitter account. The agent successfully identified what Bodhi does (a minimal content protocol with bonding curve monetization), who it's for (creators, communities, developers), and its key differentiator (no protocol fees, permissionless, permanent on Arweave), but could not resolve ambiguities about V2, practical fee mechanics, or direct competitor comparisons because those details simply don't exist in any accessible public form.

• ›Steps [11] and [19] (zhangluyao.com blog posts) were the only content-bearing sources that explained Bodhi's core mechanism, philosophy, and first application. These were discovered via web search, not from bodhi.wtf itself, and the agent explicitly cited them in the final response's Sources section.
• ›Step [12] (RetroPGF Hub) provided project classification and minimal context; step [20] (X/Twitter) provided only profile metadata. Both were cited in sources but contained limited substantive detail compared to the blog posts.
• ›Bodhi.wtf itself returned identical HTML across all routes ([1], [2], [3])—a React shell with no Server-Side Rendering, no meta descriptions, no static content, and no visible navigation or documentation. The API endpoint ([4]) confirmed the service is running but provided no human-readable information.
• ›The agent had to infer Bodhi's purpose, pricing, and positioning entirely from external interpretations by the creator (Luyao Zhang's blog) and third-party project catalogs. No competitive positioning, fee structure details, or V2 roadmap exist in any fetched or searchable public source, leaving key questions unresolved.
• ›The site is maximally agent-unfriendly for discovery-evaluation tasks: no llms.txt, no structured metadata, no public documentation, no comparison table, and no blog or help section. An agent cannot answer 'what does this do?' from the domain alone.