Philipp Krenn

EP 32 The AI Agent Trust Gap: Bridging Risk to Reliability | Elastic’s Philipp Krenn

• Elasticsearch quietly powers search across much of the internet. Philipp Krenn notes that Wikipedia and Stack Overflow run Elasticsearch behind their search boxes, and that almost everything on GitHub is cached or powered by Elasticsearch in the background — usage most people never see. Episode takeaways
• MCP is changing how systems are accessed, not just what data they hold. Krenn frames it as a shift in interaction mode: rather than writing against one specific REST API, you let the LLM figure out the MCP connection, fetch the right data, and see what actions it can run — including descriptive asks like “build me a Kibana dashboard.” Episode takeaways
• Elastic does not build large language models. Krenn says the models Elastic builds are for inference and re-ranking, and it relies on partners to supply the LLM that generates the answer — or, in Galileo’s case, the Luna models on the evaluation side. He describes the stack as an “AI lasagna” of layers where each player partners with the others. Episode takeaways