Newcomer Pod The Startup That Crashed Its Own Servers Because of Open Claw
8 snips
Mar 31, 2026 
Han Wang, CEO of Mintlify — builder of AI-readable docs and self-updating knowledge for companies like Anthropic. Jesse Zang, CEO of Decagon — maker of AI agents for contact centers and voice/chat CX. They discuss doc-driven agents, a traffic-crashing OpenClaw crawl that revealed agent behavior, voice vs chat in customer experience, safety and memory for personalized conversations, and the infrastructure needs for scalable AI-driven support.
AI Snips
Chapters
Transcript
Episode notes
Write Docs That Explain Intent Not Just Code
- Prioritize explaining intent and product context, not just code reference; docs should tell how the company thinks the product should be used.
- Han notes code-derived docs are useful but usually lack the contextual guidance agents need to choose correct behavior.
Keep Docs Self Updating To Prevent Agent Errors
- Make docs self-updating and accurate because agents use them as actionable context; stale docs lead to wrong automated answers.
- Han cites Lovable changing pricing and leaving docs outdated for a week, causing costly incorrect agent responses.
Docs Traffic Is Rapidly Shifting Toward AI Consumption
- Han estimates AI traffic to docs rose from 15% at the start of 2025 to roughly 50% today, projecting a 90/10 AI-human split soon.
- He frames docs as becoming 'knowledge infrastructure' vital to agents and chatbots.