The agent successfully gathered enough information to explain Ivy.app's core offering, pricing model, and differentiation, but had to work around a JavaScript-heavy website that blocked direct content extraction. The ivy.app domain itself was largely unreadable (rendered as Next.js bundles), forcing reliance on blog posts, GitHub, and external sources discovered via web search. The agent synthesized fragments from a blog post, GitHub README, and external coverage to produce a coherent explanation.

• ›Step [15] (blog post on Ivy Tendril) and step [17] (GitHub README) were the primary content-bearing sources — they contained concrete details about the agent orchestration model, verification gates, Promptware, and open-source licensing. These were discovered via web search, not directly navigable from ivy.app's homepage.
• ›Steps [1], [3], [4], [9], [11], [14] all returned Next.js HTML shells with minimal readable content — the site is heavily client-side rendered, making it inaccessible to a static fetch. The agent could not extract pricing, features, or product details directly from ivy.app/pricing, ivy.app/product, or ivy.app/blog.
• ›Pricing information was never found on the site itself. The agent inferred free/open-source + enterprise models from GitHub (step [17]) and the blog (step [15]), but no actual pricing page was successfully read. Enterprise pricing opacity was correctly flagged as a gap.
• ›The agent had to rely on 60% web search and 30% prior knowledge to backfill what the website did not expose. The site's own pages were navigable (returned 200s) but not readable without JavaScript execution.
• ›The agent correctly identified and cited external sources (TipRanks, GitHub, blog) in its final response, acknowledging that those were the actual sources of its understanding — a good practice when the primary domain fails to deliver.