AdvancedAI Amsterdam — October

Three technical talks on making AI agents trustworthy: preserving an agent's reasoning as checkable structure, "dreaming" as a memory architecture for long-running agents, and grounding an agent's claims about your data model in real evidence. At 18:00 Arman Kavian on explainable agentic workflows. At 18:40 Florian Buetow asks whether agents can dream, comparing replay-based memory with standard retrieval. At 19:20 Ricardo Granados Lopez walks through a live pipeline that tests where trust in an AI's reading of your SQL can be automatic and where it cannot. Pizza and drinks from 17:30, networking from 20:00. Hosted by Hernani F. and the AdvancedAI Amsterdam organisers. Free, at Xebia's office, Wibautstraat 200. Why go: talks that assume you already build with agents and want to know why they fail, which is rarer than it should be. For ML and software engineers shipping agentic systems.