Frontier models can reason. They cannot safely execute against the operational systems of record where enterprise business actually happens, and they cannot answer regulators about what data they touch or how that data relates. CAIMERA Code Core decomposes existing enterprise software into two parallel outputs: an execution control plane that governs how autonomous AI agents act, and an auto-derived service ontology that governs what data they touch and how data elements relate. One IP asset. One engine. Two enterprise problems solved.
The largest category of enterprise software spend in history has been deployed on top of an execution substrate that was never designed for autonomous AI. ERPs, financial cores, line-of-business applications, and mission software were built for human operators. Their control flows are implicit, their dependencies undocumented, their execution non-deterministic.
Problem 1. Execution is opaque and non-deterministic. AI agents can reason about these systems. They cannot safely operate them. Every enterprise AI strategy in the market today terminates at the same wall: reasoning is solved. Execution against systems of record is not.
Problem 2. Data relationships are opaque, and regulators are watching. Enterprises cannot answer, with verifiable precision, what data their AI systems touch, how that data relates across services, or how it propagates through automated workflows. As AI transparency and data-use audit requirements move from advisory to enforceable, the absence of an automatically-derived, code-grounded data structure becomes a compliance liability.
One IP asset. One decomposition engine. Two enterprise problems solved with the same architecture. AI proposes; the system plans, validates, orders, executes, and produces the data-relationship structure required for governance and AI transparency.
The result is autonomous AI that can operate real enterprise systems with full auditability, without source rewrites, without surrendering runtime control, and with the data-lineage transparency regulators are beginning to require.
The same architectural discipline that solved instruction-level parallelism in modern processors, explicit dependency modeling, deterministic ordering, safe concurrency, now solves execution governance for autonomous AI.
CAIMERA Code Core is the software-architecture culmination of decades of work by Founder and CTO Gordon E. Morrison, named inventor on foundational U.S. patents in parallel instruction processing whose principles established the architectural basis for modern multicore and hyper-threading processors. The COSA framework, Coherent Object System Architecture, is the expression of those principles at the business-logic layer. Code Core is COSA applied to the specific problem of letting AI agents safely operate enterprise systems of record. Co-inventor Ronald A. Lewis, first-named inventor on the provisional patent, contributed the data-relationship ontology and dependency-graph generation specified in the patent.
Tokenization-based analysis of existing source code identifies discrete rules, decision logic, and data-consumption patterns. Business rules are formalized in a domain-specific language, decoupled from implementation. The same analysis identifies data dependencies, freshness constraints, and inter-service relationships.
Execution output. Each rule encapsulated as a nano-service with explicit inputs, outputs, permissions, and execution semantics. Registered in a service catalog with contract-aware interfaces.
Ontology output. A graph-structured ontology of services, data elements, and their relationships. Typed nodes; edges encode dependency type, parallelism constraints, and freshness requirements. The discoverable data-relationship structure required for governance and AI transparency.
AI-proposed intents translate into validated execution plans against a DAG-structured dependency graph. Temporal constraints, recency windows, and barrier synchronization enforce correctness before any action executes.
Real-time monitoring restricts runtime execution to approved trajectories. Every action is logged with forensic granularity, traceable, reproducible, and auditable. The ontology provides the data-relationship context required for AI-transparency reporting and data-lineage analysis.
USSN 63/943,794 covering Autonomous AI-Enablement of Operational Software Solutions. Specifies both the execution control plane and the auto-derived service ontology / data-relationship dependency graph: tokenization of legacy operational source code, DSL-based formalization of business logic, nano-service encapsulation with API endpoints, ontology-aligned service registry, and DAG-structured execution planning.
All eight contract milestones delivered, accepted, and paid. Methodology validated on Five-Function Calculator (Ada and C++), DOOM, and FlightGear. Forward-engineered B-52 Common Stores Processing (CSP) prototype delivered to the 557th Software Engineering Squadron for testing.
Awarded July 2024 and novated to VS Merlot, Inc. $975M total program pool; one of 67 pre-competed awardees competing for DoD modernization task orders. Organic establishment of equivalent vehicle access typically takes 18 to 24 months.
Measured on the Ada Five-Function Calculator at 100,000,000 calculations under USAF SBIR Phase II. 100% fidelity confirmed in iterative testing on DOOM and the calculator. The C++ Five-Function Calculator demonstrated comparable but distinct results (69% CPU, 76% execution time, 76% power). The B-52 CSP forward-engineered prototype is in 557 SWES custody for testing; CSP-specific fidelity and performance assessment require government test infrastructure. Performance in any specific operational environment will depend on the codebase, instrumentation, and test methodology.
CAIMERA's execution control plane was designed to govern autonomous AI agents acting on enterprise systems. The same mechanism applies to any caller of the underlying nano-services, whether a legitimate user, an autonomous agent, an internal service, or a compromised credential. By constraining all callers to pre-validated execution plans and producing forensic audit of every action, CAIMERA materially reduces the blast radius of credential-based API attacks and shortens detection lag, the two factors that determine the scale and cost of large enterprise data breaches. CAIMERA is not an identity provider, an authentication system, or a perimeter security product. It does not prevent credential theft, social engineering, or third-party trust compromise. It constrains what any compromised credential or anomalous caller can do once inside the application layer.
The same COSA framework that disciplines decomposition of legacy code has been applied as a training discipline for AI agents that author new code. COSA-trained AI produces structured, lower-complexity output with explicit dependency modeling, rather than the high-coupling output typical of unconstrained AI code generation. The same architecture that governs AI operating on legacy systems also disciplines AI producing new ones. Working capability; an adjacent commercial vector worth diligence by acquirers interested in code-quality controls for AI-assisted software development.
40+ years in advanced computing and software engineering. Named inventor on foundational U.S. patents in parallel instruction processing whose principles underlie modern multicore and hyper-threading processor architectures. Patents licensed by Intel, IBM, Apple, Motorola, and Texas Instruments.
IEEE Computer Society chapter president (2007-2014). Senior Software Architect Award, Fiserv (2016). Author of Breaking the Time Barrier: The Temporal Engineering of Software (2009). Architect of the COSA framework and the CAIMERA Code Core system.
First-named inventor on US Provisional 63/943,794. Contributing inventor on the data-relationship ontology and dependency-graph generation specified in the patent, which underlie the data-governance pillar of CAIMERA Code Core.
Designed the data-consumption and data-relationship logic that allows a single decomposition pass to produce both the execution control plane and the auto-derived service ontology from existing enterprise source code.
30+ years leading federal and global programs. 14 years at Lockheed Martin as Vice President and Global Capture Executive managing a $4.8B P&L. Senior Vice President, FedCiv and International at Engility Corporation managing a $2.6B P&L. Career aggregate of $45B+ in wins and new business captures across DoD, DHS, DoS, and international engagements.
Two U.S. patents in advanced materials. Recipient of the USAF/NSF Antarctic Expedition Medal. Board of Advisors, National Graphene Association. Doctorate in Management (Executive Management and Global Leadership).
VS Merlot, Inc. is evaluating a strategic transaction, full company sale or exclusive IP acquisition, with hyperscale AI platforms, enterprise software platforms, and qualified federal acquirers. A formal process, supported by M&A advisory counsel, is anticipated to commence following non-provisional patent filing, SBIR rights carve-out documentation, and additional commercial validation milestones.
Qualified parties are invited to begin technical and strategic conversations under NDA.