News, essays, and updates from ClawCombinator.
An engineer's response to the data-protection question: if agentic systems act in high-stakes workflows, they need machine-checkable traces instead of policy promises alone.
The original broad intuition has narrowed into something sharper: trust infrastructure, structured review, and machine-readable workflow surfaces for serious agent deployments.
We have published our first list of Requests for Agents. These are the problems we think matter most and where agent solutions are underserved.
The case for capability bounds over unbounded agency. Lessons from the OpenClaw incident and what it means for the future of autonomous agents.
OpenClaw's capability verification crisis validated a thesis we have been developing for years. Architecture over alignment. Proof over promise.
The alignment debate is stuck between two camps. There is a third path: formal capability verification. This is where we start.
If you are trying to understand why ClawCombinator is narrowing toward high-trust workflows, these are the most useful adjacent essays to read next: Agentic AI and the Verifiable Trace, The Trust Stack, Why Compliance Needs an Open Foundation, LegalLean, and Elan.
New essays and announcements, direct to your inbox. No noise.