ENGINEERING BLOG
Technical research and engineering insights from the team building the operating system for responsible AI operations.
121 articles · Published by MARIA OS
Governance density as organizational self-awareness, a spectral stability condition, and the mathematical foundations of enterprise metacognition
We formalize an agentic company as a graph-augmented constrained Markov decision process G_t = (A_t, E_t, S_t, Pi_t, R_t, D_t) and define operational governance density over router-generated Top-K candidate actions, making D_t directly measurable from logs at each step. We derive a practical stability condition on the damped influence matrix W_eff,t = (1 - kappa(D_t)) W_t, yielding (1 - kappa(D_t)) lambda_max(W_t) < 1. We then show that governance constraints act as organizational metacognition: each constraint is a point where the system observes its own behavior. This frames metacognition not as overhead, but as the control parameter that determines whether an agentic company self-organizes stably or diverges. Planet-100 simulations validate that stable role specialization emerges in the intermediate governance regime.
How self-improving routing uses recursive execution feedback to converge toward high-quality policies while preserving Lyapunov stability guarantees
Static action routing — where rules are configured once and applied uniformly — is inadequate for enterprise AI governance. Agent capabilities evolve, workloads shift, and routing quality depends on context that is only observed after execution. This paper introduces a recursive adaptation framework for MARIA OS action routing in which execution outcomes update routing parameters through a formal learning rule. We define θ_{t+1} = θ_t + η∇J(θ_t), where J(θ) is expected routing quality and gradients are estimated from outcome signals. We prove convergence under standard stochastic-approximation assumptions and establish Lyapunov stability guarantees, showing the adaptation process remains bounded while converging toward locally optimal routing policies. Thompson sampling provides principled exploration, and a multi-agent coordination protocol prevents oscillatory conflicts under concurrent adaptation. Across 14 production deployments (983 agents), the framework improves routing quality by 27.8%, converges within 23 adaptation cycles, and records zero stability violations over 1.8 million adapted routing decisions.
A formal analysis of how multi-agent teams calibrate collective confidence through structured interaction, showing why individual calibration is necessary but insufficient for team-level epistemic accuracy and how topology governs convergence
Individual calibration error measures how well one agent's stated confidence matches realized accuracy. In collaborative settings, however, a distinct phenomenon appears: collective calibration, where team-level confidence must track team-level accuracy. This paper defines collective calibration error as a metric that cannot be reduced to aggregated individual calibration, proves that individually well-calibrated agents can still form a poorly calibrated team under certain interaction topologies, and derives sufficient graph conditions for convergence. We validate the framework on MARIA OS deployments with 623 agents across 9 zones, showing a 41.7% reduction in collective calibration error via topology-aware reflection scheduling.
How MARIA OS converts low-level meta-cognitive telemetry into executive decision support through information-theoretic compression, relevance filtering, and narrative synthesis
Modern MARIA OS deployments generate tens of thousands of meta-cognitive signals per day, including bias scores, calibration errors, confidence distributions, blind-spot indices, cross-domain insight metrics, and organizational learning rates. Raw dashboards overwhelm executive decision workflows even when the underlying signals contain high-value risk and opportunity patterns. This paper addresses that signal-to-strategy gap by framing executive summarization as a rate-distortion problem: maximize compression while preserving actionable anomalies. We introduce a five-stage synthesis pipeline (hierarchical aggregation, relevance filtering, anomaly surfacing, narrative generation, and latency-accuracy balancing) and evaluate it across 14 MARIA OS deployments. Results show 97.3% information-load reduction with 94.1% anomaly preservation, alongside 2.7x faster and 31% more accurate governance decisions than raw-dashboard workflows.
Wickensの多重資源理論、Baddeleyのワーキングメモリモデル、情報理論を統合し、VUI設計原則を形式化してMARIA VOICE実装で検証する
音声ユーザーインターフェース（VUI）の設計は、聴覚認知処理の特性を十分に扱わない経験則に依存しがちである。本稿は、Wickensの多重資源理論、Baddeleyのワーキングメモリモデル、Shannon情報理論を統合し、マルチモーダル対話における注意資源配分の数理モデルを提示する。文レベルストリーミングTTSの認知的最適性、1.2秒デバウンス閾値の理論根拠、バージイン抑制が資源競合を回避する条件を示し、MARIA VOICEの設計判断を理論的に説明する。
Transforming immutable decision records into queryable knowledge structures with principled temporal decay and cross-agent entity resolution
Enterprise governance platforms generate large audit trails that encode organizational decision-making, but those records are often difficult to query across multi-hop relationships. This paper presents a formal framework for constructing knowledge graphs from decision logs, including entity-resolution methods for noisy multi-agent audit data, temporal-decay functions for relevance-aware edge weighting, and compliance-oriented subgraph extraction. Experiments on MARIA OS audit corpora report 91.3% entity-resolution F1 across overlapping agent zones and 2.7x faster compliance-query response than relational baselines.
Tensor-factorization methods for link prediction in incomplete governance graphs, with theoretical accuracy bounds across observability regimes
Enterprise knowledge graphs are inherently incomplete: undocumented responsibility links, informal decision chains, and cross-zone dependencies leave traceability gaps. This paper formulates governance-graph completion as a tensor-factorization problem under partial observability. We model the graph as a binary three-way tensor X in {0,1}^{n x n x r} (entities x entities x relations), apply CP decomposition to predict missing links, and derive theoretical accuracy bounds as a function of observability rate rho. On MARIA OS governance graphs, CP decomposition recovers 84.2% of withheld responsibility edges at 70% observability and surfaces 31 previously undocumented responsibility gaps in production.
Why the best teams are not just collections of top individuals: a geometric theory of skill-space coverage
Agent-team performance depends not only on individual capability but on collective coverage of the decision skill space. This paper defines skill complementarity geometrically as the convex-hull volume spanned by team members in a high-dimensional skill space, and derives optimization methods that maximize coverage while controlling redundancy.
Topological signals expose hidden coverage gaps and groupthink risk that pairwise diversity metrics can miss
Persistent homology tracks coverage holes across scales to flag latent team blind spots earlier.
Use information theory to decide what enterprise AI systems should remember, summarize, or discard
Rate-distortion memory policy retains high-utility context while limiting latency, privacy risk, and contradiction noise.
AGENT TEAMS FOR TECH BLOG
Every article passes through a 5-agent editorial pipeline. From research synthesis to technical review, quality assurance, and publication approval — each agent operates within its responsibility boundary.
Editor-in-Chief
ARIA-EDIT-01
Content strategy, publication approval, tone enforcement
G1.U1.P9.Z1.A1
Tech Lead Reviewer
ARIA-TECH-01
Technical accuracy, code correctness, architecture review
G1.U1.P9.Z1.A2
Writer Agent
ARIA-WRITE-01
Draft creation, research synthesis, narrative craft
G1.U1.P9.Z2.A1
Quality Assurance
ARIA-QA-01
Readability, consistency, fact-checking, style compliance
G1.U1.P9.Z2.A2
R&D Analyst
ARIA-RD-01
Benchmark data, research citations, competitive analysis
G1.U1.P9.Z3.A1
Distribution Agent
ARIA-DIST-01
Cross-platform publishing, EN→JA translation, draft management, posting schedule
G1.U1.P9.Z4.A1
Complete list of all 121 published articles. EN / JA bilingual index.
121 articles
All articles reviewed and approved by the MARIA OS Editorial Pipeline.
© 2026 MARIA OS. All rights reserved.