Comp AI Raises $34M to Expand AI-Powered Security and Compliance Platform
Why monitoring an AI agent is not the same as controlling it — and what $34M of Series A funding says about the next compliance problem.
Read the story
Anthropic and OpenAI Say They'll Embed Independent AI Safety Evaluators
What "embedded" evaluation actually means, what evaluators are asking for, and why skepticism remains high.
Read the storyTopical takes
On AI, agents and security.
Shorter pieces first published on Substack and LinkedIn.
01AIWhen Software Starts Exercising Engineering JudgmentLinkedIn02AgentsMy agents don't know the other agents exist. That's the part that should scare you.Substack03SecurityClaude Breached Three Real Companies. Days Later, It Wiped a Database Too.Substack04SecurityOpenAI Just Proved Why AI Agents Need a Kill SwitchSubstack05CostWhy AI Agent Token Costs Are So Hard to ControlLinkedIn06CostYour token bill is not an AI problem. It is the same infrastructure trap from 2019.LinkedIn07AgentsThe web was built to exploit you. Agents don't care.LinkedIn08AgentsI studied how humans make decisions. Agents have none of those weaknesses.Substack09SecurityThe Composio Breach Exposed the Hole SOC 2 Was Never Built to SeeLinkedIn10SecurityYour SOC 2 certification says nothing about your agentsSubstack11ToolingClaude Code forgets everything when you close it. Someone built a fix and open sourced it.Substack
The newsletter
The weekly roundup, without the noise.
One email a week: the handful of stories that actually change a decision, plus the new podcast episode.
Read by 600+ engineering leaders from leading global organisations.