Launch HN: Agnost AI (YC S26) – Extract user feedback from agent conversations

Based on 5 sampled comments from this thread. Analyzed 2026-07-15.

• Praise: There's a strong belief that conversation logs can reveal hidden product signals that traditional observability overlooks — think insights from frustrated users that can guide feature development.
• Praise: The challenge of managing large datasets is being tackled with LLMs and embeddings, showing that innovative methods can still yield results even in complex scenarios.
• Praise: Creative solutions like profanity monitoring in AI setups are being highlighted as valuable learning tools, proving that even budget-conscious developers find ways to innovate.
• Concern: Skeptics argue that relying on these tools might create selective datasets, raising concerns about the validity of insights drawn from such analyses.
• Concern: There's a sentiment that the extraction of user sentiment from agent conversations is overly complicated, with worries that the AI might misinterpret user feelings.
• Concern: Commenters express frustration that traditional SQL and keyword searches can miss nuanced user feedback, suggesting that the proposed solutions may not sufficiently address core issues.