The agent successfully assembled a comprehensive explanation of Baseten by combining direct site fetches with web search results and prior knowledge. The site's heavy JavaScript rendering made full content extraction difficult, forcing the agent to rely on search results and external sources (Google Cloud Blog, Spheron, eesel AI, RunPod) to fill in pricing details, competitive positioning, and target audience. The core information was ultimately available but fragmented across the site and external coverage.

• ›Steps [1], [2], [6], [8], [9] fetched Baseten's own pages (home, pricing, about, blog, resources) but returned truncated JavaScript-heavy HTML that didn't contain readable content; the agent had to rely on web search results and prior knowledge to extract actionable details about what Baseten does, pricing tiers, and differentiators.
• ›Steps [11], [12], [13], [15] (web searches) were the actual content-bearing sources: they surfaced Google Cloud Blog partnership claim (225% cost-performance), third-party pricing breakdowns, competitive comparisons (Fireworks, Together AI, RunPod, Modal, Replicate), and audience positioning (ML engineers, data scientists). Baseten's own site did not transparently surface these comparisons.
• ›The site appears to use a Next.js frontend that renders dynamically; pricing page ([2]) returned HTML but likely requires client-side JavaScript execution to display actual rates. The agent had to work around this by citing external pricing sources (AISO Tools, Spheron Blog) rather than native page content.
• ›Target audience and use-case clarity came primarily from external reviews (eesel AI, Contrary Research) and search snippets, not from the site's own positioning. The agent noted the site does not have a clear feature comparison matrix or transparent Pro/Enterprise pricing.
• ›The agent identified legitimate gaps: Truss tool explanation, VPC vs. Cloud trade-offs, and hidden Pro/Enterprise pricing were confusing or missing—evidence the site leaves technical differentiation and advanced features under-documented for automated agents.