TimBerglund.com

Why Agent Security Breaks at the Enterprise Door — Eno Thereska on Trent.ai

Eno Thereska—formerly of Confluent, AWS, and Veeam—joins Tim Berglund to talk about his new startup Trent, which is building security for the agentic era. The conversation moves from why 70% of security problems happen before the first line of code is written, to what changes when data and compute get fused inside an LLM, to the awkward reality that AI-native startups are now hitting a wall the moment enterprises run their agents through a security review. From there it turns into a wide-ranging discussion on developer experience right now: the joy and the exhaustion of running five agents at once, the gap between shipping something fast and actually understanding it, and how trust gets built (and faked) in a world where the marginal cost of content is approaching zero. The back half is a working session on DevRel itself—how Developer Advocacy compares to the Field CTO role, when an early-stage startup should make its first DA hire, where that person should come from, and how to differentiate "the agent running on my laptop" from "the agent running inside an enterprise on enterprise data." Topics covered: * Why security feels like an afterthought in most engineering orgs * The paradox of LLMs fixing old vulnerabilities while opening new attack surfaces * Joy vs. exhaustion in agentic development * Why trust can't be shortcut, even when content can * When to hire your first Developer Advocate, and what profile to look for * Reframing "agent" for enterprise audiences who only know coding agents Guest: Eno Thereska, founder of Trent Host: Tim Berglund, https://timberglund.com