The agent assembled a reasonably complete understanding of PostMath by combining web search results and fetched company pages, but had to work around a Korean-only site with no dedicated pricing page and fragmented information across multiple domains. The site itself was navigable but not agent-optimized: core information like pricing, feature boundaries, and competitive positioning had to be reconstructed from news articles, investor profiles, and scattered references rather than published directly on the product site.

• ›The homepage (step [5]) and company page (step [20]) were successfully fetched but returned only partial product information in Korean; the agent relied heavily on web search results (steps [3], [10], [11], [16], [22], [23], [25]) for substantive details about business model, pricing, and differentiators—these search results cited news coverage (WOWTALE, VentureSquare), investor platforms (Crunchbase, PitchBook), and YouTube, not the company site itself.
• ›Pricing was the most opaque element: the agent found a reference to '500 KRW per question' and a revenue figure of 55.6 billion KRW, but no public pricing page, tier structure, or feature-to-cost mapping existed at postmath.info or mathsecr.com. The agent had to infer a freemium + usage-based model from scattered references.
• ›The site operates two domains (postmath.info as corporate site, mathsecr.com as product site) without clear explanation of their relationship; both required separate discovery. Navigation paths like /pricing and /features returned 404s, forcing the agent to rely entirely on search rather than site crawling.
• ›Core differentiators (HWP editability, LDM search technology, handwritten recognition, PenClass smart-pen system) were mentioned across company page snippets and news articles but not systematically presented on the site itself; the agent had to assemble these from fragments.
• ›The agent could not verify claimed user numbers ('100 schools' vs. '20,000 instructors') and noted inconsistencies in the public record, indicating the site does not publish clear, canonical metrics.