Research programs
A program organized
around the loop.
From algorithms to experiments, our work asks how recursive systems can improve models, search, and eventually physical understanding.
Programs
Connected research, explicit status.
Filter the active computational program from exploratory and planned directions. Visuals are conceptual unless explicitly attached to a published result.
SIML / Recursive Systems
Adaptive memory, representations, and prediction error developed as mechanisms that can be tested—not assumptions that must be believed.
View programOptimization & Search
Budgeted search across complex spaces, evaluated against credible conventional baselines.
View programSimulation Frameworks
Structured computational runs, provenance, repeatability, and comparative analysis.
View programScientific Machine Learning
Learned representations connected to physical constraints, uncertainty, and interpretable models.
View programSimulation-to-Reality
Model mismatch, calibration, and future learning from measured physical disagreement.
View programMaterials Research
Functional polymers, compliant systems, sensing, and actuation as long-term application domains.
View programEvaluation
How we measure progress.
Success is not visual novelty. It is a better model, a more useful decision, or less wasted computation under a fair comparison.
evaluations required to reach a defined target
prediction error with uncertainty reported
robustness under noise and model mismatch
useful physical knowledge per experiment
SIML competes with baselines.
Random search, Bayesian optimization, conventional active learning, and domain-specific optimizers are reference methods—not straw men. Budgets, inputs, and evaluation criteria should remain comparable.
Read the evaluation approachReal project evidence
One published computational benchmark.
These figures and values come from the existing FEPGate + V-JEPA 2 simulation report. They are not physical materials results and should not be generalized beyond that reported setup.
SIML FEPGate + V-JEPA 2
A one-bit surprise gate was added to the planner without retraining. The public research artifact includes the method, implementation link, and stated setup.
Research record
Artifacts, methods, and historical work.
Current results are separated from older conceptual material. Historical context is part of the record, not hidden in fine print.
One bit to rule the planner: SIML FEPGate + V-JEPA 2
We added a one-bit surprise gate from SIML to the frozen V-JEPA 2 planner. No retraining. Same CEM/MPC budget. ~50% lower error, ~2× faster decisions, and ~2× lower energy.
Artificial Intelligence is for Amateurs: SIML, Free Energy Principle and Cognogenics as the Foundations Toward the Birth of True Artificial Life
An early SIML paper connecting memory-constrained agents, the Free Energy Principle, and Cognogenics.
SIML Demo: Optimizing AI for Energy and Efficiency
The original SIML demo showing embodied agents adapting through memory, surprise minimization, and internal evolution.