Projects

  • The loopmath.ai home page

    loopmath.ai

    loopmath

    RSI for the masses. A command-line tool your orchestrator calls around each agent task. It compares agent workflows, prices each one before the task, and learns from every run after it.

  • The home page of the OCP documentation

    orchestrationcontextprotocol.io

    Orchestration Context Protocol (OCP)

    An open JSON format for agent orchestration runs. One document records one run: the work graph, every attempt with its harness, model, reasoning effort and token cost, and each outcome with the evidence.

  • The getviberia.com home page

    getviberia.com

    Viberia

    Civilization meets Claude Code. A desktop app for macOS and Windows that runs a team of AI coding agents from a map, the way a strategy game runs its units.

Smaller projects

  • The opening of A Statistical Analysis of LLM Benchmaxxing

    Research code

    eci-benchmaxxing

    The code and data behind A Statistical Analysis of LLM Benchmaxxing. It adds a model-level term to Epoch AI's Capabilities Index, to estimate how much a model over-performs on particular benchmarks relative to its general capability.

  • The title of The Switch

    Explainer

    The Switch

    A one-page illustrated explainer of the 2026 Nobel Prize in Physiology or Medicine.

  • Overview of the QM Raid board

    Hackathon project

    QM Raid

    An RTS-style board for QM agent swarms, built for the YC (Own Your Intelligence) Hackathon on September 27, 2026.

  • The Pelican Bike title screen

    Browser game

    Pelican Bike

    My take on the pelican riding a bike. Excitebike + Pelicans.

  • The dagr pane: a run's graph of attempts and gates

    herdr plugin

    herdr-dagr

    A herdr pane that shows an agent swarm's plan as a live graph.

  • Accuracy of the early takes, set by set, from the Lords of Limited post

    Python tool

    yt-pundit-analyzer

    A tool that checks how well pundits' predictions held up. It takes a video of early takes and a later video on what actually happened, pulls the key points from each with Gemini, and writes a comparison. It was built for the post on Lords of Limited's early format guesses.