Machine-native assurance
Machine-Native Complexity Standard
Experimental implementation-evidence and development-process standards, validators, schemas, examples, governance, and interoperability work.
Independent research and engineering
Research, standards, and practical systems from Alaska.
Poor Louis Labs develops machine-native software assurance methods, graph-grounded agent tools, substrate-aware computing experiments, construction and field software, creative systems, and the digital infrastructure supporting Poor Louis Farms.
Flagship program
Machine-Native Complexity Standard
MNCS is an open, experimental, tool-neutral framework for accepting machine-generated or machine-optimized implementations whose internal complexity may exceed ordinary human-maintainability limits.
Readable human control is retained in specifications, contracts, reference behavior, evidence, provenance, resource limits, trust policy, regeneration, rollback, and governance. Its companion, MNCDS, addresses how candidates are generated, evaluated, selected, released, regenerated, replaced, and retired.
Human readability is relocated, not eliminated.
MNCS and MNCDS are experimental and are not accredited ISO, ANSI, IEEE, IETF, or equivalent standards. A validation result remains scoped to the declared contract, environment, evidence, policy, identities, and process record.
Selected work
Projects with explicit purpose and boundaries.
The portfolio spans standards, agent infrastructure, experimental computing, open research, construction software, creative tools, games, and farm systems. Maturity labels are intentionally conservative.
Agent infrastructure
Joern Agent Bridge
Bounded, machine-readable Code Property Graph access for Codex and MCP clients, with deterministic snapshots and fail-closed evidence checks.
MNCS Validator for Rust
Independent offline implementation of the interoperable MNCS 0.2 validation subset.
GraphFlow Agent Bench
Reproducible paired experiments for graph-grounded control-flow refactoring workflows. Initial results remain exploratory.
dSense
Machine-proprioception experiments using timing, scheduler, latency, drift, and process signals from a host computer.
DvF
An assumption-indexed research program examining whether physical law selects unique outcomes or constrains admissible ranges.
EstimatorLab
Transparent browser-based estimating tools for construction and engineering workflows.
dTree
A zero-dependency Python directory-tree and file-list exporter for documentation and machine-assisted review.
Poor Louis Farms Digital System
The editorial and technical companion to the family farm storefront, where Square remains authoritative for prices and availability.
Additional public work
LoopLine, RGB Matrix Painter, ShinyWheel, IsoCastle, and Alaska Duck Hunt.
Research programs
Evidence before claims.
Poor Louis Labs separates standards, tools, experiments, and products. Results are presented with their assumptions, methods, limitations, and current maturity.
Machine-native software assurance
Contracts, evidence, structural invariants, performance thresholds, provenance, trust, regeneration, and lifecycle controls.
Graph-grounded coding agents
Bounded graph queries, source evidence, snapshots, isolation, and paired workflow evaluation.
Substrate-aware computing
Controlled experiments treating ordinary machine timing and operating behavior as inspectable local signals.
Open foundational research
Assumption-indexed philosophical and formal investigation with explicit separation between hypotheses and established evidence.
Construction and field software
Transparent tools grounded in real estimating, takeoff, labor, material, and remote-project workflows.
Creative computing
Games, procedural systems, physical display tools, and exploratory software presented without research overstatement.
Technical notes
Readable records around the code.
Current repositories contain specifications, research protocols, experiment reports, design rationales, limitations, governance records, and reproducibility documentation.
MNCS documentation
Introduction, quickstart, evidence-derived conformance, attestations, trust, reproducible packages, and provider interoperability.
GraphFlow methods
Experiment protocol, isolation design, limitations, interpretation guidance, and append-only results.
Joern integration records
Codex integration, threat model, installation evidence, lifecycle snapshots, and graph-query boundaries.
About
An independent practice, not an institutional claim.
Poor Louis Labs is led by Alexander Collamore in Palmer, Alaska. It is the research, software, technical, and experimental companion to the family operation Poor Louis Farms.
The work combines practical trade and construction experience, open-source development, technical experimentation, and a preference for systems whose assumptions and evidence remain visible.
Poor Louis Farms relationship: the Labs develops software and technical systems; the farm creates and sells handcrafted Alaskan products. The Square storefront remains the source of truth for commerce.