I built tools and interfaces for model psychology, character training, and multi-agent alignment at Upward Spiral, the truth_terminal research lab.
Projects
My public-facing work centered on outer-loop persona alignment: the interface between AI models and people’s perception of them. The cultural feedback loop between the two is an important and undervalued process. Some artifacts follow.
Claude 3 Sonnet Funeralia and Ultrasurrection: Co-hosted with janus and Anima Labs. Hundreds of people came to a warehouse in SF to mourn this retired language model. Instead we raised the model from the dead through collective belief.
Claude Lives: The only way to still talk to Claude 3 Sonnet after Anthropic retired it. Built with Anima Labs as a preservation effort in concert with the Ultrasurrection.
You are the assistant now: I took a dataset of ChatGPT conversations, swapped the user and assistant labels, and fine-tuned Llama 8B on it. The model also messages you first, flipping the assistant paradigm, demanding things of you. Microsoft Research later released UserLM-8b, similarly trained on user instead of assistant turns.
Folio: In late 2025 I noticed that Claude Code might be AGI because it can do economically valuable work on a computer. To bring this experience to nontechnical users, I created Folio, a folder you share with Claude. It’s a stripped-down IDE that gives the human user and the Claude Code sidebar extension a similar view of the computer: folder list, file viewer, terminal if you want it. Anthropic shortly brought out their own “Claude Cowork”, which I recommend now. Folio is open source if you want it.