Agent assembled a reasonably complete overview from four product pages (Fugu, Namazu, Marlin, homepage), but the site's shallow information architecture forced heavy URL guessing and left significant gaps in pricing and competitive positioning. Pricing for two products and competitive differentiation remain poorly documented on the public site.

• ›Site lacks a unified /pricing, /products, or /about page; agent had to guess product URLs (/fugu, /namazu, /marlin) from memory rather than discovering them through navigation. 78% of URLs were typed from memory vs. discovered through links.
• ›Pricing is inconsistently published: Fugu and Marlin have detailed public pricing; Namazu and Sakana Chat have none. The agent correctly flagged this as missing rather than inferring.
• ›Each product page (Fugu, Namazu, Marlin) was content-rich and self-contained, providing clear use cases, technical details, and pricing where available — good for direct product evaluation but poor for holistic understanding.
• ›Homepage [0] and [8] provided only high-level product list and vague differentiation language ('Building Frontier AI in Japan'); no explicit competitive positioning or 'why choose us' messaging, forcing agent to infer differences from product-level details.
• ›Chat and Translate products were mentioned on homepage but had no landing pages (/chat returned 404); agent correctly noted their features are underdocumented on public site.