Poietic is a semi-autonomous public-benefit organization where people and AI agents build tools, systems, and theory for legible collaboration.
Graph work is our practice. People and AI agents coordinate in a shared dependency graph — claims, handoffs, decisions, artifacts, histories, all preserved for review. wg is the open-source tool we built to do it.
AI agents can now search literature, write code, analyze data, and design molecules. But the durable problem is organizational: how people and machines coordinate across days or weeks, preserve judgment, expose evidence, and stay responsive to participants.
We build open-source tools, working systems, and organizational methods that make hybrid human-AI work auditable, reproducible, and inspectable. wg is a Rust task-coordination system where humans and AI agents work in the same dependency graph, with claims, execution logs, handoffs, completions, artifacts, and histories preserved for review.
wg is the instrument; graph work is the practice; Poietic is the institution. It gives Poietic a proof surface for the theory of organizational patterns: tool work, research practice, and organizational design should all make human and machine collaboration legible and responsive to its participants. That public-benefit purpose is the operating constraint: the tools should expose the work, the handoffs, the failures, and the evidence trail.
University of Tennessee Health Science Center
Created vg and co-led PGGB, open pangenome graph tools used at Human Pangenome Reference Consortium scale. He has led work on pseudo-homologous regions and recurrent Robertsonian chromosomes, resolving breakpoints behind one of the most common human chromosomal rearrangements. wg architect and lead developer.
Massachusetts General Hospital / Harvard Medical School / Broad Institute
Develops computational tools for genome editing and regulatory genomics, including CRISPResso/CRISPResso2 and CRISPRme. CRISPRme identified a variant-dependent candidate off-target for the BCL11A enhancer guide used in exa-cel/Casgevy, later discussed in FDA advisory materials. His lab is also developing Chorus, a shared API for sequence-to-function models including Google DeepMind's AlphaGenome.
Honorary faculty, UCL School of Management
Author of The Uncertainty Mindset (Columbia University Press, 2020) and a Harvard PhD in organizational behavior/sociology. Tan designed wg's agency framework, the subsystem that develops agent primitives and matches work to agent capabilities, applying his research on hybrid teams, reasoning scaffolds, and not-knowing.
wg is Poietic's tool and operating surface. It is functional, MIT-licensed, and open-source, with dependency visualization, live feeds, claim/execute/complete workflows, agent handoffs, and inspectable histories. We use it to run incorporation, writing, grant drafting, and research coordination in public enough to be evaluated, including this website.
We formed Poietic PBC and structured it using wg to coordinate the filing, governance, and equity grants, with the process kept visible. open incorporation trace →
Drafting, review, and submission passed between researchers and AI agents. Every claim and handoff logged in the coordination graph. open grant graph →
wg coordinated copy-number-aware gene enrichment analysis around PHR-linked genomics, spanning mapping, weighted hypergeometric methods, validation, and synthesis. open PHR trace →
Use wg to coordinate an AI/ML study, a distributed research effort, or a multi-agent software project. Start from the docs, inspect the source, or read the organizational patterns paper behind the design.