Agentome
Structure-preserving Large Physics Models (LPMs) form the core. Continual personalization connects incoming BCI data to backend automation and frontend agents that support each user’s goals.
Agentome · Continual Large Physics Models
Build on full stack platform for a classical–quantum hybrid digital twin, powered by structure-preserving Large Physics Models (LPMs) that track neural states and forecast neural-field dynamics, with continual personalization from incoming brain–computer interface recordings — to support learning, focus, meditation, music, art, and restoration, alongside mental-health and neurological research workflows.
Conceptual state-model illustration
Not a measured result
The core and its applications
Agentome connects a core family of Large Physics Models to two agent layers. Three proposed programmes explore how personalized models and guided experiences could support users and researchers.
Structure-preserving Large Physics Models (LPMs) form the core. Continual personalization connects incoming BCI data to backend automation and frontend agents that support each user’s goals.
A planned frontend guide for learning and focus, informed by personal neural-state estimates, task history, and user goals. Test lesson pacing and break suggestions against independently measured learning outcomes.
Explore guided meditation, personalized music, creative art, sleep, and restoration. Frontend agents would combine user preferences with evaluated model outputs to propose sessions for study.
Support investigators with longitudinal records, model comparisons, and evidence-linked hypotheses for mental-health and neurological research, using independently assessed study outcomes.
The Agentome experience
Our Large Physics Models (LPMs) describe neural dynamics. Continual LPMs (cLPMs) are the planned personalized versions updated from incoming recordings, with evaluation before an update is accepted.
Manage recording quality, data preparation, training, evaluation, deployment, and monitoring within agreed settings. The aim is for users to take part without manually operating the modeling infrastructure.
Help users set goals, understand model-informed suggestions, choose sessions, and give feedback. Guidance spans learning, focus, meditation, music, art, and restoration; research workflows retain investigator review.
Where the work stands
Core model research includes an internally reported state-model prototype. Continual personalization and both agent layers need model-specific and workflow-specific evaluation. Application studies remain proposals.
One structure-preserving LPM family for neural-state tracking and neural-field forecasting, with model-specific evaluation.
Backend orchestration for data and models; frontend support and application guidance. Integrated workflows are in development.
Nine proposed use cases across Superlearning, Restoration, and Mental Health & Neurological Research.
A precise modeling scope
We use “digital twin” to mean a fitted model of recorded dynamics, tied to its input recordings and model version. A comparison across sessions must account for recording quality, acquisition changes, and task context.
Development milestones
These are development stages, not launch dates. Each stage calls for a concrete research artifact or a predefined comparison.
Internally reported
Document state tracking and neural-field forecasting separately, including the internally reported prototype and its evaluation scope.
Next evidence milestone
Document the dataset, preprocessing, baselines, held-out split, metrics, and failure cases in a reviewable benchmark report.
In development
Connect continual model updates, backend operations, and frontend guidance. Evaluate update acceptance, deployment rollback, tool errors, and the usefulness of explanations.
Proposed studies
Test learning and wellbeing use cases against task-history policies or fixed guidance using predefined outcomes.
Collaboration sought
Evaluate cross-visit model comparisons with acquisition controls and independently assessed study outcomes.
Collaborate with Mindverse
Describe your recordings, study question, and current analysis baseline. We can discuss a scoped comparison and the evidence it would need.