Inside the Agent Engine: A LangChain and Traversal Fireside Chat
Traversal builds SRE agents that search petabyte-scale logs with no labeled training data
“a single investigation's probably gonna cost you the GDP of a small country”
LangChain CEO Harrison Chase hosted Traversal co-founders to discuss building AI agents for production incident troubleshooting, revealing why SRE is one of the hardest agentic domains: no labeled data, LLMs untrained on telemetry, petabyte-scale data that makes naive agent search economically unviable, and zero tolerance for wrong answers under stress. Traversal's approach centers on a 'production world model' built offline from logs, code, and Slack to give agents system context before an incident occurs. The conversation is a useful real-world stress test of current agent limitations at enterprise scale, relevant for anyone designing agents that must search large, fragmented, unstructured data sources.