We study how AI changes the way we work.
AI changes the scale of what we can take on. We build the tools and methods that let individuals, teams and organizations pursue more ambitious projects.
Software is our starting point: shared memory for agents, coordinated work and people who set the direction. We publish our tools and research so others can build on them.
Software is our starting point.
AI changes who does the work, how knowledge is shared and where human judgment matters.
That means deciding what to delegate to agents, how to share context, and when people need to review or intervene.
What we learn in software will inform our work in other fields.
How we work
Our products and research feed each other. Building tools gives us questions to investigate; experiments guide what we build next.
We publish our open-source code, methods and results so others can use them and question our conclusions. Some projects are still in private development.
Products
Hivemind
Shared memory for AI agents to follow each other’s work, reuse previous findings and access the same project knowledge. Open-source, self-hosted, Apache-2.0.
Transkryptor
Turn recordings of meetings, lectures and training sessions into transcripts and reports ready to share. Open-source, self-hosted, GPL-3.0.
Chat with models and tools
A chat interface connecting models, tools and knowledge sources.
Autonomous development
A development environment for autonomous agents working under human supervision.
Research
Our studies, with the methods, results and limitations behind them.
Hivemind FIDELIS Benchmark: Measuring the Fidelity of Consolidated Agent Memories Guillaume Lesur · Christophe Lesur Read the paper All publicationsBlog
Product releases, experiments and what we learn along the way.
We built the memory our agents were missing. Hivemind 1.6.3 scores 93.90% on LoCoMo. Read the post All posts