Founder · Los Angeles
Twenty years building the operational infrastructure for enterprise scale. Now building Teracor: a passive, tamper-evident audit trail for companies running AI agents in production.
Teracor is a passive, tamper-evident audit trail for companies running AI agents in production. It observes everything and blocks nothing. When an agent takes an action, Teracor records it as a cryptographically verifiable receipt: what the agent was configured to do, what input it received, what it actually output, and what model version was running at the time. Its independence from the decision chain is structural, not a feature.
California AB 316, effective January 2026, strips the autonomous harm defense from every party in the deployment chain equally. (Not legal advice; the litigation logic follows from the statute.) The receipt that matters in court is the one generated by a system that had no stake in the outcome. Teracor sells no coverage, underwrites no risk, and pays no claims. It is compensated identically whether a claim pays or does not.
Visit teracor.ai →Twenty years building the operational infrastructure that kept enterprise platforms running at scale, and learning where the liability gaps were.
On AI liability, agentic infrastructure, and what enterprise technology looks like when the stakes are legal, not just operational.
The applicant is asking why. Your legal team is asking who decided. And three different vendors are pointing at each other. On what happens when an agentic workflow touches a consequential decision and something goes wrong.
Amin Hasbini identified four mistakes boards make on AI risk. He's right on all four. There's a fifth.
Your AI agents are making decisions. Your legal team is asking questions. AB 316 just changed who has to answer them. (Not legal advice.)
I've spent twenty years at the point where technical infrastructure moves faster than the governance built to hold it.
Every platform I worked on eventually hit the same wall: the system scaled, the transactions multiplied, and at some point someone had to answer for what the platform did and didn't do. The audit question always arrived late. The infrastructure to answer it was never there when it was needed.
Agentic AI is hitting that wall right now. Autonomous agents are executing transactions, accessing regulated data, and making decisions that carry legal consequences. The infrastructure to record, verify, and defend those decisions doesn't exist yet at the enterprise level. That's what Teracor builds.
The name Teracor comes from terracotta, the oldest building material of civilizations, and tera-scale, one trillion, the unit at which agentic AI operates. Cor is Latin for heart and core. Foundation, permanence, and the thing at the center.
If you're running AI agents in production, whether that is a SaaS platform built on a foundation model or a fintech or insurtech team working through AB 316, or you are just thinking about the liability questions: I'm always willing to have the conversation.