01
Governed AI & agent authorization
AI systems with explicit permissions, pre-action controls, human review points, and records that distinguish a proposed action from an executed result.
Evidence: COSMIC-ALLOW · T2Helix · Project Epistemic Bound
AV Family Enterprise LLC · Veteran-owned · Pennsylvania
Applied intelligence.
Accountable systems.
An independent, veteran-owned research and development company working across scientific software, artificial intelligence, computational modeling, and research infrastructure.
We build instruments for understanding complex systems — and infrastructure for keeping the resulting work accountable. A hypothesis is not a result, a prototype is not a deployed system, and an AI-generated proposal is not an authorized action.
Fig. 01 · Software
Fig. 02 · Science
Fig. 03 · Record
- type
- Instrument
- lifecycle
- Living · v1.21.0
- evidence
- In use, self-verifying
- type
- Study
- lifecycle
- Closed · v4.3
- evidence
- Bounded null
Veteran-owned
Founded and led by a U.S. Army veteran
Pennsylvania-based
Registered Pennsylvania LLC
Founder-led
Scoped engagements with direct accountability
Evidence-first
Every claim carries its type, lifecycle and evidence
01 / Capabilities
Six areas of work, one method
Each area turns a difficult question into inspectable models, bounded experiments, and documented findings. Projects appear beneath the capability they substantiate — not the other way round.
02
Persistent knowledge & research infrastructure
Software architectures for preserving research context, tracing claims to evidence, recording corrections, and supporting continuity across people, models, sessions, and software.
Evidence: Sovereign Stack · public chronicle · temple-harness
03
AI evaluation & human–AI workflows
Research tools and assessment workflows for examining model inputs, system behavior, evidence quality, human oversight, and the limits of an evaluation’s conclusions.
Evidence: The Black Box Boundary · co-creation case study · Master Operations Agent
04
Local-first AI & architecture R&D
Experimental AI architectures, locally operated inference workflows, and comparative studies of model behavior under practical computing constraints.
Evidence: Phenomenological Compass · Phase-Modulated Attention
05
Scientific modeling & simulation
Computational models, interactive scientific instruments, synthetic scenarios, and experimental software for examining complex system behavior.
Evidence: entropy campaigns · operator training simulator · Temple Lab
06
Computational biology & pharmacology research
Mechanistic modeling, literature-linked hypothesis development, and computational analysis in support of experimentally testable biological questions.
Evidence: VDAC1 Pharmacology Atlas · Gate-Jamming Score
02 / Selected evidence
Three branches, three kinds of standing
A published protocol, a shipping training instrument, and an untested hypothesis set. They are not the same kind of thing, and the labels say so.
Illustration · Protocol invariants
ground state → deny
allow → requires an unforgeable sentinel
boundary → kernel sandbox floor, not the policy layer
Governed AI · RFC · Published v1.1 · Supported within scope
COSMIC-ALLOW
A cross-vendor proof-of-allow protocol for AI coding-agent cockpits. Deny is the ground state; allow requires positive, unforgeable proof; the sandbox floor is the load-bearing boundary. Reference implementation in cosmic-cli (Apache-2.0).
Limits
Not a claim that any specific deployment is secure.
Synthetic training environmentSimulation · Independent training aid · v3.1.0 · Shipping
Operator training simulator
A browser-based environment for exploring synthetic process behavior, alarms, trends, operator controls, and instructor-driven scenarios, with training records and a standalone offline build.
Limits
Not a Honeywell product. No certified qualification or site connection implied.
Scientific illustrationComputational biology · Hypothesis set · Awaiting wet-lab validation · Untested in vivo
VDAC1 Pharmacology Atlas
Mechanistic hypothesis development for mitochondrial gating: a cofactor model of how hexokinase-II, Bcl-xL, and cholesterol jointly set the apoptotic threshold, with 24 testable predictions and a candidate biomarker.
Limits
None of the 24 predictions has been tested at a bench. Two of three cohorts returned nulls.
Also in the portfolio
- Sovereign Stack · instrument, in use
- Phenomenological Compass · study, supported within scope
- The Black Box Boundary · study, bounded n=1
- Entropy as a Tunable Equilibrium · study, bounded null
- Project Epistemic Bound · instrument, release blocked
- Master Operations Agent · research workbench, ongoing
03 / Working method
What are you trying to understand? What instrument will help? What evidence would justify trusting the result?
Every engagement answers those three questions before it answers anything else. Methods, limitations, and corrections stay part of the work.
Step 1
Define the problem
Agree what useful success looks like, what constraints matter, and what remains uncertain.
Step 2
Bound the system
Specify data access, actions, human approvals, failure handling, and what the prototype is not authorized to do.
Step 3
Build the smallest useful version
Prefer a testable, understandable solution over unnecessary complexity or scale.
Step 4
Test against evidence
Use appropriate scenarios and controls, record actual outcomes, and retain failures and unresolved questions.
Step 5
Hand over recoverable work
Deliver documentation, versioned artifacts, and enough context for another competent person to review or continue the project.
04 / Founder
Anthony J. Vasquez Sr.
Founder and managing member · U.S. Army veteran · Independent researcher
Anthony brings a practical, hands-on perspective to software and AI research, with an emphasis on accountability, useful tools, and disciplined technical follow-through. His wider public research — spanning governed AI, dynamical systems, AI architecture, and computational pharmacology — is published through The Temple of Two.
About Anthony, the research origins, and the standards behind every claim →Bring a defined problem. Build a reviewable path forward.
Government organizations, research institutions, and industry partners are welcome to discuss scoped software development, computational studies, evaluation tools, and research collaborations.
Technical project
A defined prototype, evaluation tool, simulation environment, or evidence-management workflow — with agreed scope, acceptance criteria, data handling, documentation, and handoff.
Research collaboration
A question, the work already completed, the unresolved limitation, and the expertise or validation needed next.