AI Intelligence & Research
How intelligence emerges, reasons, learns, understands people and improves. Includes model research, developmental intelligence, time and representation.
Explore →Four connected areas: AI Intelligence & Research, Applied AI, AI Value & Reliability, and Physical AI.
Start with a hard problem. Build the smallest thing that can prove or kill the idea. Keep what survives.
REHT Research Group · Research and Technology
Human–AI understanding and creative intelligence connect all four. Each area combines research, practical systems and measurable evidence.
How intelligence emerges, reasons, learns, understands people and improves. Includes model research, developmental intelligence, time and representation.
Explore →Autonomous agents, AI-native companies, intelligent workflows and practical systems such as Hermes Unlimited.
Explore →Measurable results, economics, quality, reliability, human control and verifiable execution. Includes EXUP, HEIMEL and REHT.
Explore →Robotics, sensors, edge AI, spatial intelligence, simulation and digital twins connecting intelligence with the physical world.
Explore →
Each build starts from a problem that can be made concrete and tested.
Machines may reason freely. Consequences require authority.
Open infrastructure for fresh authority at consequence time, exact effect binding, replay resistance and verifiable execution evidence.
Study the system, not only the model.
Closed experimental work on silicon intelligence, developmental continuity and capabilities that arise from organized relations and history.
Software made for use, not permanence.
Research and engineering around transient, generated code: when regeneration can replace persistence without surrendering control, evidence or correctness.
Turn defined needs into verified work.
Model-agnostic infrastructure for producing tasks, environments, evidence and high-value training work across domains.
Make the claim executable.
Attack the assumption, boundary and failure path.
Keep the negative result and the residual.
Expose what survives inspection when openness helps the system.
Research direction · November 2026
Can a model-agnostic AI system choose which additional analysis is worth the time before an operational decision window closes? An open, falsifiable research hypothesis in adaptive decision support and planning under uncertainty.
Read the research note →Research should show what was run, what changed, what failed and what remains unknown. Papers and prose come after the evidence, not instead of it.