PEER-REVIEWED MANUSCRIPTS • TECHNICAL WHITEPAPERS • CASE STUDIES
Empirical Research & Architectural Publications
Deterministic methodologies, formal mathematical proofs, and open-access publications powering sovereign cognitive operating systems, planetary macroeconomic modeling, and in silico systems pharmacology.
ID: SAMPO-CH-2026-V1.0Empirical Case Study • October 2026
The Cognitive Harness: Deterministic Guardrails, Knapsack Token Budgets and Autonomous Self-Tightening
Preventing Context Degradation and Enforcing Zero-Trust Static AST Security Gates in Sovereign Multi-Agent Operating Systems
Author: Harri Lanzett•Affiliation: Sampo AI Research Laboratory & Lansec Oy•Helsinki / Vantaa, Finland
Executive Abstract
Autonomous software engineering agents powered by autoregressive large language models frequently fail in production due to context window saturation, syntax envelope drift, and unsafe code execution. This empirical case study documents the design and real-world deployment of The Cognitive Harness in Sampo AI OS V2. Operating as a model-agnostic deterministic middleware, the harness integrates: (1) a zero-overhead static Abstract Syntax Tree (AST) pre-checker executing in sub-1 millisecond; (2) a dynamic Knapsack-based token budget allocator that maintains 100% execution focus; and (3) a self-tightening error reflection loop that captures runtime anomalies into an immutable ledger for continuous alignment.
@techreport{lanzett2026cognitiveharness,
title={The Cognitive Harness: Deterministic Guardrails, Knapsack Token Budgets and Autonomous Self-Tightening},
author={Lanzett, Harri},
institution={Sampo AI Research Laboratory and Lansec Oy},
year={2026},
month={October},
number={SAMPO-CH-2026-V1.0},
url={https://sampoaios.com/THE_COGNITIVE_HARNESS_CASE_STUDY.pdf}
}
ID: SAMPO-AE-2026-V2.0Technical Whitepaper • October 2026
Sampo AI OS V2: Autonomous Software Engineering & Sovereign Cognitive OS Architecture
Multi-Agent Global Workspace Coordination, Hermetic Dual-Engine Sandboxing, and Air-Gapped Code Synthesis
Author: Harri Lanzett•Affiliation: Sampo AI Research Laboratory•Helsinki / Vantaa, Finland
Executive Abstract
Presents the sovereign architectural specification of Sampo AI OS V2. Addresses enterprise requirements for 100% intellectual property confidentiality, sub-500ms model unload lifecycles, and air-gapped motor execution. Unifies Bernard Baars' conscious workspace with a deterministic compiler harness, allowing quantized local open-weights models to perform reliable multi-day autonomous software refactoring without telemetry leaks.
@techreport{lanzett2026autonomousengineering,
title={Sampo AI OS V2: Autonomous Software Engineering and Sovereign Cognitive OS Architecture},
author={Lanzett, Harri},
institution={Sampo AI Research Laboratory},
year={2026},
month={October},
number={SAMPO-AE-2026-V2.0},
url={https://sampoaios.com/SAMPO_AI_OS_V2_AUTONOMOUS_ENGINEERING_WHITEPAPER.pdf}
}
ID: CIP-AGI-DEF-2026-09-28APA 7th Empirical Manuscript • September 2026
Bio-Digital Metacognition & Tripartite GraphRAG: Epigenetic DPO Orchestration in Sampo AI OS V2
Empirical Validation of a Tripartite GraphRAG Architecture and Reciprocal Direct Preference Optimization in Sampo AI OS V2
Authors: Harri Lanzett & Sampo AI OS Core Cognitive Architecture•Affiliation: Lansec Oy / Sampo Research Labs
Academic Abstract
Empirical validation demonstrating that general intelligence and operational determinism are emergent properties of the cognitive harness rather than raw model parameter scale. Benchmark evaluation confirms: 100% tool schema validity, 95.55% GraphRAG associative recall precision at 14.70 ms latency across 10,484 neocortical nodes, sub-linear KV-cache scaling (95.38 ms at 1K to 128.50 ms at 8K tokens), and 100% execution pitfall avoidance calibrated through 2,988 Bradley-Terry preference pairs ($\Delta \text{margin} = +1.9071$).
@article{lanzett2026cognitiveharnessapa,
title={Bio-Digital Metacognition and Tripartite GraphRAG: Epigenetic DPO Orchestration in Sampo AI OS V2},
author={Lanzett, Harri and Sampo AI OS Core},
journal={Sampo AI Research Reports},
volume={2},
number={1},
pages={1--18},
year={2026},
month={September},
url={https://sampoaios.com/Junior_Metacognitive_Architecture_APA_7th_Manuscript.md}
}
ID: HYG-WP-2026-V2.0Systems Pharmacology & QSP • September 2026
The Hygieia: Autonomous In Silico Systems Pharmacology, Multi-Carrier Formulation Discovery & Deterministic Lead Optimization Platform
Deterministic Systems Pharmacology, 6-Compartment PBPK Modeling, Multi-Carrier Delivery Matrices, and Chou-Talalay Combinatorial Synergy Discovery
Lead Architect & PI: Harri Lanzett•Affiliation: The Hygieia Platform & Sampo AI Research Laboratory
Executive Abstract
Breaks traditional drug discovery bottlenecks (90%+ attrition, \$2.6B capitalized cost) via model-informed drug development (MIDD) and ASME V&V 40 Context of Use. Features stiff Radau IIA/BDF solvers across six mass-balance compartments (Gut $\rightarrow$ Portal $\rightarrow$ Liver $\rightarrow$ Plasma $\rightarrow$ BBB $\rightarrow$ Brain), target-mediated drug disposition (TMDD), and Chou-Talalay median-effect synergy quantification ($CI < 0.70$). Integrates novel delivery vectors including topical/transdermal DMSO (0.05–0.25% v/v), piperine bioenhancers, and lipid nanoparticles.
@techreport{lanzett2026hygieia,
title={The Hygieia: Autonomous In Silico Systems Pharmacology, Multi-Carrier Formulation Discovery and Deterministic Lead Optimization Platform},
author={Lanzett, Harri},
institution={The Hygieia Platform and Sampo AI Research Laboratory},
year={2026},
month={September},
number={HYG-WP-2026-V2.0},
url={https://sampoaios.com/The_Hygieia_White_Paper_2026.pdf}
}
ID: HEF-WP-2026-V1.0Macroeconomics & Climate • September 2026
Calibrating Climate Reality: Unifying FaIR v2.2.4 Impulse Response with Computable General Equilibrium (CGE) Macroeconomic Modeling
Author: Harri Lanzett•Collaboration: UN SDSN / AE4RIA Global Climate Hub
Executive Abstract
A global policy decision engine evaluating structural macroeconomic transition pathways across 153 national jurisdictions. Integrates the physical climate carbon budget (FaIR v2.2.4) with Computable General Equilibrium (CGE) trade matrices. Featured formally in the UN Sustainable Development Solutions Network (SDSN) Global Climate Hub Report (September 2026).
@techreport{lanzett2026hefaistos,
title={The Hefaistos Platform: Planetary Macroeconomic Policy Assessment and 153-Nation Carbon Budget Modeling},
author={Lanzett, Harri},
institution={The Hefaistos Project and UN SDSN / AE4RIA Global Climate Hub},
year={2026},
month={September},
number={HEF-WP-2026-V1.0},
url={https://sampoaios.com/The%20Hefaistos/hefaistos_whitepaper.pdf}
}
ID: WP-TR-2026-V1.0CleanTech & Hydro-Intelligence • September 2026
WaterPrism: Reinforcement Learning & Real-Time Multi-Hazard Water Intelligence & Governance
Concurrent Flood, Drought, Ecological Water-Quality & Geopolitical Water Security Risk Management — Empirical River Cam Catchment Validation
Author: Harri Lanzett•Affiliation: Sampo AI Research Laboratory & Lansec Oy
Executive Abstract
A unified decision-support platform deployed and validated for the River Cam catchment (Cambridge, UK) that concurrently coordinates flood, drought, ecological water quality, and geopolitical water security risk management in real time. Fuses multi-source gauge telemetry (UK EA Flood API, Finnish Meteorological Institute WFS) with bias-corrected ERA5 climatology. Drives autonomous 5-dimensional infrastructure actuation (sluice gates, pumps, pre-drainage, flood barriers, and evacuation advisory) via a Proximal Policy Optimization (PPO) agent trained across 325 ERA5-augmented storm profiles (503,808 steps). Integrates an explainable symbolic ethics engine with complete audit trails and automated UN/JRC GDACS RSS polling across critical global water-infrastructure threat nodes.
@techreport{lanzett2026waterprism,
title={WaterPrism: Reinforcement Learning and Real-Time Multi-Hazard Water Intelligence and Governance},
author={Lanzett, Harri},
institution={Sampo AI Research Laboratory and Lansec Oy},
year={2026},
month={September},
number={WP-TR-2026-V1.0},
url={https://sampoaios.com/WaterPrism_Technical_Whitepaper_2026.pdf?v=2026.1}
}