Скачать или смотреть Vjaceslavs Klimovs on Why 40% of Security Work Lacks Threat Models

Vjaceslavs Klimovs, Distinguished Engineer at CoreWeave, secures the infrastructure powering AI training and inference at massive scale. He explains why detection should come before prevention in fast-moving environments, how AI agents will expose previously tolerable gaps like bearer tokens, and CoreWeave's approach to host integrity from firmware through user space.

Chapters:
0:00 Introduction
0:29 Security journey and background
1:33 Zero to one at CoreWeave
2:25 Security observability as foundation
4:44 Detection-first approach and forensics
6:12 Threat-model-driven security work
13:00 Why security work lacks threat models
18:10 Silicon Valley innovation problem
21:11 AI agents in security teams
26:26 Design docs and LLM efficiency
28:10 LLM strengths in security domains
34:08 Investigation space with AI
35:59 Detection as code imperative
41:03 Three pieces of advice