Agent successfully extracted and synthesized what Haga does, who it serves, and how it differs from alternatives by fetching the homepage three times and assembling a coherent narrative. The site publishes its core positioning, methodology, and founder/team status in fetchable form, but does not expose pricing details, active customer deployments, or complete case studies—gaps the agent appropriately flagged rather than fabricated.

• ›All three fetches were to the same URL (homepage typed from memory, not discovered via site navigation), yet returned different response summaries suggesting the agent performed logical extraction and reorganization of a single page's content across multiple retrieval attempts. This indicates the homepage is content-dense and agent-readable but not hierarchically structured with separate URLs for pricing, team, or methodology sections.
• ›Pricing information is genuinely absent from the site: the agent did not fail to find it, but correctly identified that Haga operates on custom-quote basis with no published rate card. This is a real gap, not a navigation failure.
• ›The site explicitly names competitors (Antioch, Bifrost AI, Instance, Robocurve) and articulates differentiation (unified world-model + policy verification), suggesting competitive positioning is intentionally surfaced for evaluation tasks.
• ›Founder stage is transparent: Mushood Hanif is named, technical stack is disclosed, and active hiring is stated. The agent correctly inferred 'early stage' from these signals rather than guessing.
• ›Technical credibility is high-signal on the page: specific detector performance metrics (recall 1.0, zero false positives), named tools (CoTracker3, Physics-IQ), and documented failure modes are all published—making the agent's confidence in the methodology grounded in disclosed facts, not model memory.