The agent successfully gathered enough information to explain Superhuman's core business, pricing, and differentiation, but relied heavily on web search and prior knowledge to fill gaps left by the website's JavaScript-heavy rendering. The site's main pages (homepage, about, mail, docs, plans) loaded but provided incomplete pricing breakdowns and bundle details, forcing the agent to supplement with external sources and infer positioning from search results.

• ›Step [1] (homepage) and step [14] (plans page) returned structural HTML but pricing details and feature comparisons were embedded in JavaScript and not rendered in the fetched response, leaving the agent unable to extract bundle pricing or full tier comparisons directly from the site.
• ›Step [4] (about page) established the company rebrand and ownership structure (Grammarly → Superhuman holding Docs/Mail/Grammarly/Go), but the site itself did not clearly disambiguate what 'Superhuman' refers to (company vs. Mail product) — the agent had to discover this confusion through search steps [15] and [17].
• ›Steps [11], [12], [15], [17] (web searches) were the primary sources for pricing ($30–$40/month for Mail), competitor comparisons, and the recent rebrand timeline. The agent had to exit the domain entirely because the site did not publish these comparisons or clearly articulate its market differentiation.
• ›Step [7] (Mail page) and step [9] (Docs page) provided product descriptions but no feature details; step [14] (plans page) existed but failed to render pricing and feature matrices in a machine-readable form.
• ›The site navigation structure (separate URLs for /mail, /docs) suggested modularity, but the site did not clarify subscription models (bundled vs. separate subscriptions, licensing options). The agent had to flag this as 'confusing/missing information' in the final response.