Humans and machines on an architecture that retains knowledge and structures decisions — researched in the open: from AI-assisted knowledge work to safeguarding public spaces, without biometric identification.
We research in the open — and for selected companies we build systems they own afterwards. Work with us
Proof, not claims
What we publish, what we release as code, and where the architecture is in use.
Preprint on arXiv (2604.23878) and open records with DOIs on Zenodo.
Five works cite ZenBrain (as of 28 September 2026): a survey by the Sico team at Microsoft Research, MindMemOS by Huawei's Noah's Ark Lab, the EVD preprint, the HindsightTag manuscript and the Lamperouge manifesto.
ZenBrain: the open memory core on npm, Apache 2.0, 822 tests (as of 10 October 2026).
Try it in the browser: the playground runs the real library.
Who we work with
The ZenBrain architecture connects seven cognitive systems — from remembering to metacognition — into one memory.
7 memory layers — from active context to lasting knowledge, with spaced-repetition-based scheduling.
Cross-context reasoning, multi-step inference, and automatic fact-checking.
Persistent agent loops with checkpoints, tool composition, and dynamic teams. Pause, resume, debate.
PPTX, XLSX, PDF, DOCX — straight from the chat. CLI agent for the terminal.
Knowledge gap detection, hypothesis engine, information gain scoring.
Pattern tracking, prediction engine, and prediction error learning.
Confidence calibration, capability profiling, feedback loop. Calibrated uncertainty instead of hallucination.
Architecture
ZenBrain is our open-source memory core. Multiple applications build on it — the same seven-layer architecture, only the domain adaptation varies.
Memory layers
Research track — public-space safety
Research on early recognition of critical situations in public spaces and at mass-casualty incidents — with civil liberties built into the architecture by design.
Research prototype · GDPR Art. 89 · designed under the EU AI Act
Three paths to work with this research — an open-source core, open-access publications and scientific cooperation.
ZenBrain — our memory core as an open-source library. TypeScript, zero dependencies, Apache 2.0.
Explore the open core →Open-access preprint on arXiv, permanently archived with a DOI on Zenodo — including replication material and citation formats.
View the publications →Research collaborations, joint funding proposals and methodological exchange with universities, companies and the public sector.
Discuss a collaboration →Self-hosting, EU data protection, and the EU AI Act's transparency duties are built in as design constraints, not add-ons. We provide the architecture, not a certificate.
The paper, the code, and the Playground are open — check our work before you trust it. For companies that want to build their own system on it — one they own afterwards — working together is the next step.
See how we work