The agent successfully synthesized a comprehensive explanation of Runway Dev by combining fragments from multiple sources, since the main portal (dev.runwayml.com) is a client-side JavaScript SPA that returned no meaningful content. It relied heavily on external documentation (docs.dev.runwayml.com), web search results, and API documentation to answer what Runway Dev does, who it's for, how it's priced, and how it compares to alternatives. The site itself is poorly navigable for machine agents — the portal pages are non-rendering SPAs, and critical information is scattered across docs.dev.runwayml.com, help articles, and news posts.

• ›The main domain (dev.runwayml.com) failed to expose machine-readable content; all three fetches of the portal returned identical truncated HTML with only meta tags and GTM script — no actual product information, pricing, features, or docs were present in the DOM. The agent had to pivot to docs.dev.runwayml.com and web search.
• ›Step [7] (docs.dev.runwayml.com/) and step [14] (docs.dev.runwayml.com/guides/pricing) returned Astro-rendered pages that were truncated in the response, preventing the agent from extracting complete pricing tables or feature lists directly. The agent backfilled this gap using web search results (steps [11], [12]) that referenced the same docs pages.
• ›Step [9] (api-docs.runway.team/) was fetched but contributed minimally — it's a Redoc OpenAPI spec interface, which only returned boilerplate HTML structure without actual endpoint details in the response snippet. The agent did not cite it in the final answer.
• ›Web search (steps [11], [12], [15], [17], [18]) was the primary source of actionable content: pricing details, competitive positioning, target audience, and feature comparisons came from third-party aggregators (eesel.ai, lovart.ai, cybernews.com, apiframe.ai) rather than from Runway's own pages. These searches supplied 57% of the agent's information sources.
• ›The agent correctly identified and worked around the SPA limitation, explicitly noting in the final response that 'the main portal at dev.runwayml.com is a client-side JavaScript app, so I couldn't see actual UI details.' It did not attempt to run or scrape JavaScript, and instead pivoted to alternative resources.
• ›Information fragmentation: pricing is documented at docs.dev.runwayml.com/guides/pricing; models are listed at dev.runwayml.com/models (cited but not fetched); FAQs are on help.runwayml.com; announcements are on runway.com/news. The agent had to synthesize these across five different domains.