VALO RESEARCH

We build what is missing.

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

Four connected areas.

Human–AI understanding and creative intelligence connect all four. Each area combines research, practical systems and measurable evidence.

AI Intelligence & Research

How intelligence emerges, reasons, learns, understands people and improves. Includes model research, developmental intelligence, time and representation.

Explore →

Applied AI

Autonomous agents, AI-native companies, intelligent workflows and practical systems such as Hermes Unlimited.

Explore →

AI Value & Reliability

Measurable results, economics, quality, reliability, human control and verifiable execution. Includes EXUP, HEIMEL and REHT.

Explore →

Physical AI

Robotics, sensors, edge AI, spatial intelligence, simulation and digital twins connecting intelligence with the physical world.

Explore →

The work is the explanation

VALO evidence and execution artifacts
Tests before claims. Running evidence remains visible.

Current work

Each build starts from a problem that can be made concrete and tested.

01OPEN · RUNNING

HEIMEL

Machines may reason freely. Consequences require authority.

Open infrastructure for fresh authority at consequence time, exact effect binding, replay resistance and verifiable execution evidence.

02ACTIVE LAB

Synapse Lab

Study the system, not only the model.

Closed experimental work on silicon intelligence, developmental continuity and capabilities that arise from organized relations and history.

03ACTIVE RESEARCH

Useware

Software made for use, not permanence.

Research and engineering around transient, generated code: when regeneration can replace persistence without surrendering control, evidence or correctness.

04BUILDING

Factory

Turn defined needs into verified work.

Model-agnostic infrastructure for producing tasks, environments, evidence and high-value training work across domains.

Tests before claims

01

Build

Make the claim executable.

02

Break

Attack the assumption, boundary and failure path.

03

Measure

Keep the negative result and the residual.

04

Publish

Expose what survives inspection when openness helps the system.

Research direction · November 2026

Time-Critical Decision Intelligence

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 →

Running evidence, not a blog

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.