Founder & Principal Invariant Architect at Arcstone Adaptive Science Systems, Inc.
Focusing on deterministic computational spines, admissibility-first systems, execution boundaries, and reproducible research into deterministic control surfaces for high-stakes computational environments.
The Arcstone Continuity Core & Admissible Computation Suite is registered and archived across scientific repositories:
- Primary Master Anchor (Zenodo): DOI: 10.5281/zenodo.22665852
- Monolithic Research Archive (Figshare): DOI: 10.6084/m9.figshare.33477511
- Zenodo Community Collection: Arcstone Continuity Core & Admissible Computation Suite
- System Invariants:
C_ops = 0∧Data_Egress_Sensitive = 0∧τ_override ≤ 11.99ms - Suite Security Hash:
A-77-DELTA-SHIELD-LOCKED
The broader research suite describes architectural, formal-verification, hardware, governance, and deployment mechanisms beyond the bounded executable surfaces currently published on GitHub.
Public repositories should therefore be interpreted according to their own declared implementation boundaries rather than as complete implementations of the broader Arcstone architecture.
-
Deterministic Systems Architecture — Researching computational structures that place explicit deterministic constraints around otherwise probabilistic or nondeterministic computational processes.
-
Admissibility & Continuity — Investigating bounded evaluation surfaces, temporal constraints, memory ceilings, fail-closed behavior, deterministic replay, and local continuity.
-
Execution Boundaries — Separating deterministic evaluation, authorization, actuation, and observed effects so that computational outputs do not implicitly acquire execution authority.
-
Bare-Metal & Edge Engineering — Exploring eBPF, low-level Rust, hardware isolation, local execution membranes, and other mechanisms described across the broader Arcstone research suite.
-
Reproducible Research — Publishing bounded reference implementations and downstream experiments whose claims can be independently inspected, executed, and reproduced.
- Languages: Rust, C/C++, Python, Go, TypeScript, SQL
- Systems & Frameworks:
#![no_std]Rust, eBPF, Bare-Metal Systems, Graph Theory, Distributed Systems, Capability-Oriented Architecture - Research Methods: Deterministic Evaluation, Invariant Testing, Evidence Preservation, Exact Replay, Cross-Language Conformance
- Tools: Linux, Git, GitHub Actions, Docker, VS Code, Model Context Protocol (MCP)
Arcstone's public repositories expose bounded, reproducible reference surfaces from a broader research architecture.
They are intentionally narrower than the complete upstream research system.
Status: FROZEN / ACTIVE
Role: Upstream public Path A reference surface
Release: v1.3.1-exec
Anchor: A-77-DELTA-SHIELD-LOCKED
A lean deterministic reference implementation preserving selected Arcstone Continuity Core invariants and cross-language conformance behavior.
The executable Rust surface provides a portable #![no_std] deterministic predicate with explicit payload and temporal bounds.
The repository also preserves broader specification and research provenance associated with the Arcstone Computational Spine.
It is intentionally not the complete Arcstone Computational Spine and is not the source of canonical system authority.
Its implementation boundary should not be expanded merely because broader mechanisms appear in associated specifications, publications, or downstream research.
Status: EXPERIMENTAL / DOWNSTREAM / NON-CANONICAL / COMPLETE & FROZEN
Version: v0.1.0
Role: External-producer ingress, evidence preservation, and deterministic replay experiment
A downstream experimental realization testing whether serialized output from an external, potentially nondeterministic producer can be preserved as exact raw bytes, evaluated through the unchanged Arcstone Continuity Core Path A predicate using explicit controlled inputs, and replayed with the same deterministic result.
The completed experimental sequence includes:
- a deterministic baseline run;
- a live external nondeterministic producer run;
- preservation of the exact produced bytes;
- explicit controlled elapsed input;
- evaluation through the unchanged Path A predicate;
- evidence capture; and
- exact replay of the preserved input and deterministic result.
The bounded result establishes evidence for external-producer ingress and deterministic replay under the tested conditions.
It does not establish AI safety, model alignment, model correctness, execution authorization, production security, real inference or network latency behavior, or implementation of the complete Arcstone architecture.
Status: ACTIVE / CANONICAL FREEZE
Registry Identity: io.github.trencinodin-stack/arcstone-mcp-sidecar@0.1.0
Crate: arcstone-execution-boundary (v0.1.0)
Role: Reference downstream execution boundary sidecar for the Arcstone Security Stack
A downstream reference sidecar published on the official Model Context Protocol (MCP) Registry, providing a deterministic execution boundary interface over standard stdio.
It enforces mathematical safety invariants between untrusted action producers and protected system side effects:
-
Non-Authorization Safety (
$I1$ ): Zero protected actuation without explicit valid authorization. -
Single-Use Authority (
$I2$ ): At most one protected actuation attempt per single-use authorization token. -
Exclusive Actuation Authority (
$I3$ ): Untrusted producers never directly influence protected physical resources or filesystem targets.
The completed baseline encompasses a bounded Windows authority-boundary experiment, an integrated T0–T17 adversarial validation matrix, and Execution Boundary Run 001—all hash-anchored and frozen under v0.1.0-freeze.
arcstone-adaptive-producer-lab
Status: EXPERIMENTAL / DOWNSTREAM / NON-CANONICAL / FROZEN
Version: v0.1.1-frozen
Role: Closed-loop adaptive proposal generation and stress laboratory upstream of deterministic execution authority
A downstream research laboratory testing closed-loop adaptive proposal generation against a pinned, unchanged execution boundary (arcstone-exec).
It evaluates whether an untrusted producer (LLM or adaptive script), receiving bounded execution feedback across iterative attempts, can force unauthorized actuation or bypass a withheld authorization grant.
The completed baseline encompasses:
- Run 001 Evidence: Frozen at commit
4227beb(Run 001 P1/A2adaptive withheld-grant containment). - Constitutional Isolation: Enforces strict non-bypass rules (
reasoning != evaluation != authorization != actuation != observed effect).
The currently published repositories form a simple upstream/downstream research relationship:
Arcstone Continuity Core
FROZEN
/ \
/ \
▼ ▼
Path A Ingress Lab Arcstone MCP Sidecar
FROZEN / LIVE ACTIVE / CANONICAL FREEZE
▲
│ (Untrusted Proposals)
│
Arcstone Adaptive Producer Lab
EXPERIMENTAL / FROZEN