Selected work · Reviewed public inventory
One execution-authority kernel, four applied systems, and a curated collection of open tools and public experiments.
01 · Core platform
The decision engine every applied ExecLayer system is built around.
The authority layer
LiveThe decision engine that checks whether an AI action is authorized before it happens and creates verifiable proof of the decision.
Know which actions may proceed, stop the ones that may not, and retain evidence of every decision.
02 · Applied systems
Four systems carrying the same pre-execution authority model into distinct operating domains.
Powered by the ExecLayer Kernel
Active developmentApplies the authority kernel to robots and industrial systems, stopping unsafe or unauthorized machine actions before they occur.
Keep machines inside approved operating limits before motion begins.
Powered by the ExecLayer Kernel
LiveApplies the kernel to autonomous AI agents, preventing them from using tools or changing systems without passing enforceable policy checks.
Require an enforceable decision before an AI agent can use a tool or alter a system.
Powered by the ExecLayer Kernel
Active developmentApplies the kernel to autonomous financial activity, ensuring transactions cannot proceed without policy approval and verifiable evidence.
Prevent autonomous transactions from proceeding without policy approval and an audit trail.
Powered by the ExecLayer Kernel
LiveA production monitoring system for Ring devices that applies the same pre-execution authority model to real-world security workflows.
Monitor events and automate responses while keeping actions inside explicit operating rules.
03 · Open tools and experiments
Independent tools and research surfaces that can be inspected, used, or evaluated on their own.
Useful work released publicly
Active developmentA private, local tool that gives AI coding companions a voice using the Mac's built-in speech system.
Hear coding-companion responses without sending speech to a hosted voice service.
Useful work released publicly
Active developmentLets AI assistants read and interact with the modern web, including pages that require a real browser session.
Give an assistant access to browser-rendered pages through a CLI or MCP server.
Useful work released publicly
ResearchAn interactive visualization showing how different careers may be affected as AI capabilities advance.
Explore possible patterns of AI-driven job change through an accessible public visualization.
Useful work released publicly
Active developmentPrivate, persistent memory for AI assistants, allowing them to retain project context without requiring a hosted memory service.
Keep useful project memory local while giving assistants durable context across sessions.
Start with the live refusal or bring us the action your system needs to control.