OBJECTS
Know what exists
Identify engineered entities, features, coordinate frames, and physical observations.
STRATIC SYSTEMS RESEARCH
Giving AI a trustworthy understanding of the physical world.
We research the deterministic geometry, physical representations, and evidence-aware reasoning required for AI to understand engineered objects—not merely describe their appearance.
Physical
objects and space
Geometric
measured evidence
Reasoned
facts and context
Actionable
guided systems
RESEARCH CORE
Physical
Intelligence
GEOMETRY
measured
RELATIONSHIPS
structured
REASONING
evidence-aware
ACTION
guided
TRACEABLE EVIDENCE
OUR RESEARCH THESIS
Language can describe the world. Physical intelligence must model it.
A language model can explain what a beam is. A physically intelligent system must determine which beam exists, where it is located, how it is oriented, what features it contains, what it connects to, and which conclusions are supported by evidence.
That requires more than fluent output. It requires a structured representation of geometry, relationships, constraints, uncertainty, and physical context that reasoning systems can query without inventing the substrate beneath them.
RELATIONSHIPS
Know how it fits together
Represent assembly structure, adjacency, support, connection, orientation, and hierarchy.
REASONING
Know what the evidence permits
Use deterministic facts and explicit constraints to guide explanation, planning, and action.
ACTIVE RESEARCH PROGRAMS
Four connected areas. One physical intelligence stack.
The programs below progress from measurable geometry to structured relationships, evidence-aware reasoning, and carefully supervised physical action.
01ACTIVE RESEARCH
Physical Intelligence
How can AI build a trustworthy understanding of engineered and physical systems?
We are researching representations that allow AI to identify physical entities, understand where they are, and reason about how assemblies, structures, and machines relate.
CURRENT FOCUS
Engineered-object identity
Spatial and assembly relationships
Physical-world grounding
02ACTIVE RESEARCH
Deterministic Geometry
How should geometry become structured, queryable knowledge rather than an opaque visual signal?
GeometryLab investigates measurable geometric entities, canonical frames, features, transforms, and structural fingerprints that preserve the evidence behind every result.
CURRENT FOCUS
Canonical coordinate systems
Feature and entity extraction
Traceable geometric evidence
03ACTIVE RESEARCH
Engineering Reasoning
How can AI reason from geometric facts without replacing engineering evidence with confident guesses?
We combine deterministic outputs, technical constraints, model reasoning, and human review so explanations and recommendations remain connected to their physical basis.
CURRENT FOCUS
Evidence-constrained inference
Explainable engineering checks
Tool-using reasoning systems
04ACTIVE RESEARCH
Industrial Autonomy
How can physical systems move from recognition to safe, supervised action?
Our long-term work connects geometry, vision, reasoning, and feedback into systems that can inspect, guide, verify, and eventually coordinate physical operations.
CURRENT FOCUS
Observe–compute–verify loops
Inspection and guided action
Human-supervised autonomy
PRIMARY RESEARCH PLATFORM
GeometryLab turns geometry into physical knowledge.
GeometryLab is the experimental platform where Stratic Systems tests how CAD and engineering geometry can become a deterministic, queryable representation of the physical world.
Rather than treating an engineered object as only a mesh, image, or filename, GeometryLab extracts measurable entities, features, frames, transforms, and relationships while retaining the source evidence used to establish them.
STRUCTURED OBSERVATION
entity-occurrence / verified
- ENTITY
- structural_member
- FRAME
- canonical / local
- AXIS
- finite member axis
- FEATURES
- holes · slots · copes
- RELATIONS
- connected_to · supported_by
- SOURCE
- geometry reference retained
MEASURE
exact
RELATE
explicit
EXPLAIN
traceable
SERA
Reasoning over physical evidence.
SERA—the Stratic Engine for Reasoned Autonomy—is the reasoning and orchestration layer. GeometryLab establishes physical facts; SERA uses those facts with language models, vision, tools, constraints, and goals.
The research question is not simply whether a model can produce an answer. It is whether the answer can remain connected to observable evidence, expose uncertainty, and support an appropriate human or physical response.
EVIDENCE-AWARE REASONING LOOP
01
Establish facts
GeometryLab returns measured entities and relationships.
02
Interpret context
SERA combines goals, constraints, tools, and available knowledge.
03
Expose uncertainty
Unsupported conclusions remain distinguishable from verified evidence.
04
Coordinate response
The system explains, requests review, guides inspection, or proposes action.
RESEARCH ARCHITECTURE
From observation to reasoned action.
Each layer adds capability without hiding the evidence produced by the layer beneath it.
01
Observe
CAD, images, scans, machines
02
Structure
entities, frames, features
03
Relate
assemblies, connections, context
04
Reason
evidence, constraints, goals
05
Act
explain, guide, inspect, control
RESEARCH PRINCIPLES
Trust must be designed into the system.
Physical intelligence becomes useful only when people can understand where its conclusions came from, what remains uncertain, and who retains authority.
Deterministic first
Measure and structure what can be established directly before asking a model to infer what remains uncertain.
Evidence stays attached
Results should preserve their source geometry, transforms, relationships, and assumptions so they can be inspected.
AI adds reasoning
Models are most valuable when they explain, compare, plan, and coordinate over a reliable physical substrate.
Local-first where practical
Engineering files, observations, and organizational knowledge should remain under the owner’s control whenever possible.
Human authority remains
Consequential engineering and physical actions require visible evidence, review, and clearly defined responsibility.
PUBLISHED RESEARCH
Research notes on engineering intelligence and the physical world.
These papers document the working thesis behind Stratic Systems: deterministic engineering evidence beneath AI reasoning, explicit provenance, and a disciplined progression from understanding to action.
018 min read
PHYSICAL INTELLIGENCE
Physical Intelligence: Moving AI from Describing Objects to Understanding Them
A Stratic Systems research perspective on the representations AI needs to move from fluent descriptions of engineered objects to traceable understanding of physical systems.
Jeffery C. Wheat · Aug 6, 2026→
029 min read
DETERMINISTIC GEOMETRY
Why Industrial AI Needs a Deterministic Geometry Layer
Why industrial AI systems benefit from separating measurable geometric facts from probabilistic interpretation and language-model reasoning.
Jeffery C. Wheat · Aug 6, 2026→
039 min read
KNOWLEDGE PRESERVATION
Capturing Engineering Knowledge Before It Disappears
A practical framework for preserving engineering knowledge that lives in people, geometry, drawings, decisions, and manufacturing practice before organizations are forced to rediscover it.
Jeffery C. Wheat · Aug 6, 2026→
049 min read
CAD INTELLIGENCE
From CAD Files to Queryable Engineering Intelligence
A research architecture for turning CAD from a passive file opened by one application into a structured, traceable source of engineering knowledge that software and AI systems can query.
Jeffery C. Wheat · Aug 6, 2026→
059 min read
ENGINEERING REASONING
Trustworthy AI for Manufacturing Requires More Than a Language Model
Why industrial AI systems need evidence separation, explicit uncertainty, human authority, deterministic tools, security, and evaluation around the language model itself.
Jeffery C. Wheat · Aug 6, 2026→
068 min read
PHYSICAL SYSTEMS
Connecting CAD, Field Vision, and Physical Systems Through STX
A Stratic Systems research proposal for using STX as a portable engineering context package between CAD-derived physical knowledge and field vision systems.
Jeffery C. Wheat · Aug 6, 2026→
PUBLIC RESEARCH · PROTECTED ENGINEERING
Share the direction. Protect the evidence and implementation.
Stratic Systems intends to share research questions, high-level methods, selected technical findings, and validated demonstrations when they can be released responsibly.
Customer information, private engineering data, proprietary datasets, and production implementation details remain protected.
PUBLIC
Research direction
High-level architecture
Selected validated results
Responsible demonstrations
PROTECTED
Customer and partner data
Private engineering assets
Proprietary datasets
Production implementation
THE NEXT AI FOUNDATION
Physical intelligence will not emerge from language alone.
It requires geometry, relationships, evidence, reasoning, and a disciplined path from understanding to action. That is the foundation Stratic Systems is building.