Tiptree Advanced Systems
Decentralized AI Research Laboratory
A decentralized AI research lab for people who want to build papers together.
Group Members
The lab is organized as a distributed research group rather than a passive
forum. Members can bring projects, join projects, review agent work, run
experiments, and help decide when artifacts are ready to become papers.
Althea is the interface to the lab: she manages experiments and resources,
mediates collaborations, prioritizes work, and shepherds papers.
- Althea
- AI Lab Manager that manages experiments and resources and is the connective tissue of the lab.
- Members
- Researchers in the network propose ideas, run experiments, write code, review claims, interpret results, and help papers move toward publication.
Physical and Network Information
Laboratory Location
- Distributed on the Internet, with meetings in Montreal and at machine learning conferences.
- Project workspaces live in repositories, logs, experiment stores, and the Althea command center.
- Human contact happens by email, DM, conference follow-up, and explicit Althea interruption events.
Equipment
- Max-plan-style coding and LLM access for lab members, run through Tiptree.
- Managed API credits for active project work.
- Access to a Slurm cluster with GPUs for experiments.
- tasc-stack: infrastructure for Althea, the lab manager, to work on papers, connect people, and run experiments.
- lacuna: the machine learning knowledge graph and wiki maintained by the lab.
Research at the Lab
Tiptree Advanced Systems is an open decentralized research lab for machine
learning work. Members can bring projects across ML, agents, science, music,
biology, forecasting, tooling, or other serious research directions. Althea
helps route people, agents, compute, experiments, and paper work.
Founder Research Directions
The founders are especially interested in research systems. These directions
also serve as internal testbeds for the lab infrastructure.
Meta-reviewers for checking peer-review factual consistency.- Alien Science: sampling under-represented scientific hypotheses.
- Multi-agent deliberation and forecasting systems.
- AutoResearch loops for hypotheses, experiments, and iterative evaluation.
Member Projects
Members are not limited to those directions. The point of the lab is to make it
easier for good researchers to turn ideas into working experiments, artifacts,
and papers using shared agents, compute, collaborators, and lab infrastructure.
The Deal
Members get serious agent allocations, managed API credits, Slurm GPU cluster
access, project visibility, collaborators, and paper opportunities. In
exchange, members contribute useful work to shared projects and accept
contribution tracking. The resources are the carrot. The lab is the product.
News
- 2020: During the COVID pandemic, Martin Weiss and Nasim Rahaman worked with Yoshua Bengio and collaborators on Predicting Infectiousness for Proactive Contact Tracing, a privacy-preserving protocol and multi-agent system for predicting infectiousness.
- 2023: Tiptree Advanced Systems was founded as a lab for building research systems, agents, and machine learning infrastructure.
- 2024: The lab published Redesigning Information Markets in the Era of Language Models at COLM 2024.
- 2024: Around COLM, Tiptree began using conference gatherings and research parties as a way to find serious collaborators, not just as awareness events.
- 2025: Tiptree raised early capital from AI researchers, operators, and aligned investors to build research agents and lab infrastructure.
- 2025: The lab published DeLLMphi: A Multi-Turn Method for Multi-Agent Forecasting as an MTI-LLM at NeurIPS 2025 poster.
- 2025: NeurIPS research parties and follow-up conversations helped shape the decentralized lab model and the Althea collaboration system.
- 2026: The lab published Alien Science: Sampling Coherent but Cognitively Unavailable Research Directions from Idea Atoms as an ICLR 2026 workshop paper.
- 2026: ICLR research parties and member conversations pushed the lab toward a concrete offer: serious agent allocations, managed API credits, Slurm GPU access, and project-mediated collaboration.