"An agent is not a prompt. It is a system that decides, executes, and is held to account."

Skavinski / AI · 2026

Engineer the substrate autonomous intelligence is built on.

Agent Tooling

Honest Evaluation

Independent Lab

What we are building

ADE

Agent Development Environment

A first-class workspace for designing, testing, and deploying agents. Built around multi-step reasoning, tool use, and long-horizon execution.

orchestrate harness deploy

Bench

Benchmarking Suite

Proprietary evaluations for agents and frontier models. Coding, mathematics, market structure, regulatory reasoning, and our own internal models.

code math markets regulation

Also

We publish research notes, prototypes, and tutorials on agent development as the work progresses.

One update when something is real. No newsletter. No churn.

You're on the list. We'll send one update when something is real.