Home · Writing

Writing

Decision-ready writing on applied intelligence as a systems discipline: architecture, context, evaluation, cloud platforms, human authority and the deeper questions behind reliable machine capability.

GOVERNANCE PLANEidentity · policy · evaluation · human decision rights
governed contextsource · freshness · scope
bounded reasoningplan · abstain · escalate
controlled actiontool contract · receipt
verified outcomeeffect · evidence · owner

Architecture · Deployment · Cloud · Design · Research

Featured
Architecture

Deterministic-First Agents: The Certainty Gradient for Governed Banking Automation

A production architecture for decomposing banking workflows into deterministic controls, calibrated models and bounded reasoning, so autonomy is…

Architecture

Governed Context Architecture: Reconstruction, Provenance and Active Memory for Long-Horizon Agents

A bank-grade context architecture that reconstructs current state, enforces authorization before retrieval, links claims to evidence and treats…

Research

Model-Risk Evidence for Agentic Systems: From Outcome Accuracy to Trajectory Assurance

A model-risk and production-assurance architecture for evaluating an agent's complete decision path: evidence, tools, order, policy, recovery,…

Topics

Architecture

Production AI systems design

Deployment

Production controls, evaluation and operating evidence

Google Cloud

Agentic AI on GCP

AWS

Agentic AI on Amazon Web Services

Design

Interfaces and knowledge for human-agent teams

Research

Current research translated into design decisions

Machine Intelligence

First principles of representation, control and search

Consciousness

Machine consciousness research and assurance

All writing

Architecture

Production AI systems design
Agent Memory as an Attack Surface: Provenance, Quarantine, DriftMemory systems · Security and containment · Evidence and provenanceAn Agentic Test Migration Factory with Repository GatesReliability and recovery · Agent architecture · Production systemsAnatomy of a Production Multi-Agent SystemAgent architecture · Production systems · Identity and authorityContext Caches under Banking Privacy ConstraintsMemory systems · Privacy and analytics · Production systemsContext Compilers for Policy-Filtered RetrievalContext engineering · Evidence and provenance · Identity and authorityContext is the New ComputeContext engineering · Platform economics · Reliability and recoveryDecision Decomposition in Commercial Credit: Evidence, Arithmetic, Policy, Narrative and AuthorityIdentity and authority · Financial services · Evidence and provenanceDesigning Multi-Agent Topologies with Computable Blast RadiusAgent architecture · Identity and authority · Risk and regulationDeterministic-First Agents: The Certainty Gradient for Governed Banking AutomationProduction systems · Reliability and recovery · Risk and regulationEvent-Sourced Relationship Intelligence Without a Shadow Customer RecordIdentity and authority · Production systems · Agent architectureFreshness, Supersession and Valid-Time EvidenceTemporal systems · Evidence and provenance · Production systemsGoverned Context Architecture: Reconstruction, Provenance and Active Memory for Long-Horizon AgentsEvidence and provenance · Temporal systems · Production systemsGraphRAG, Relational State or Vector Search?Retrieval systems · Representation · Financial servicesLegal Document Investigation as Reconciliation, not SummarisationProduction systems · Evidence and provenance · Identity and authorityMultimodal Context for Lending and Financial CrimeFinancial services · Evidence and provenance · Security and containmentThree Principals, One Decision: Agent Identity and Delegated Authority in Regulated SystemsIdentity and authority · Production systems · Agent architecture

Deployment

Production controls, evaluation and operating evidence
Agent Reliability Atlas: 171 Production Failures, Controls and Recovery PathsProduction systems · Reliability and recovery · Agent architectureAML Alert Triage with a Hard SAR Boundary: Evidence Assembly Without Delegating SuspicionEvidence and provenance · Financial crime · Production systemsCalibrated Cascade Routing in Production: Small Models First, Confident EscalationReliability and recovery · Production systems · Platform economicsContinuous Control Evidence: From Cloud Configuration to Auditor WorkpaperProduction systems · Risk and regulation · Reliability and recoveryDrift: Behavioural Sentinels and Why Tool Selection Can Lead Output MetricsReliability and recovery · Production systems · Financial servicesKYC Periodic Review as an Evidence-Bound Workflow: Events, Provenance and Human AuthorityEvidence and provenance · Financial crime · Risk and regulationMeasuring Hybrid Retrieval Beyond Recall@kRetrieval systems · Reliability and recovery · Production systemsPopulation-Scale Outcome Testing for Customer InteractionsProduction systems · Interface design · Evaluation and assuranceRCSA Evidence Assembly Without Delegating the RatingProduction systems · Evidence and provenance · Risk and regulationReal-Time Fraud Investigation Agents: Fast Evidence, Bounded Action and Human AuthorityReliability and recovery · Production systems · Financial servicesRegulatory Change as a Traceable Obligation GraphRisk and regulation · Identity and authority · Production systemsShipping a Private Skill Model Foundry: Distillation, LoRA, Calibration, and Red Teaming Inside a VPCReliability and recovery · Production systems · Security and containmentShipping LLM Systems in Regulated EnvironmentsReliability and recovery · Risk and regulation · Production systemsThe Eval Pipeline is the ProductProduction systems · Reliability and recovery · Evaluation and assuranceThe Evaluation Stack for Agentic Systems: Golden Cases, Adversarial Scenarios, and Process EvidenceEvaluation and assurance · Security and containment · Risk and regulation

Google Cloud

Agentic AI on GCP
BigQuery as the Agent's Structured Memory: Medallion Pipelines, CDC and Row-Level SecurityAgent architecture · Cloud platforms · Memory systemsBigQuery as the Memory Layer for Agent PlatformsMemory systems · Production systems · Agent architectureBuilding Bank-Grade Agents on Agent Runtime: LangGraph and a Controlled Path to ADKAgent architecture · Production systems · Identity and authorityGemini Enterprise Agent Platform: What Actually Ships in a BankProduction systems · Identity and authority · Agent architectureRetrieval on GCP: Hybrid Dense and Sparse Search with RRF and Cross-Encoder RerankRetrieval systems · Production systems · Platform economicsVPC Service Controls and Workload Identity for Agentic EstatesCloud platforms · Identity and authority · Production systems

AWS

Agentic AI on Amazon Web Services
Agentic Architectures on AWS: AgentCore, Durable Workflows, and Where Lambda EndsCloud platforms · Agent architecture · Production systemsAmazon Bedrock AgentCore in Production: A Regulated-Industry Field ReportProduction systems · Cloud platforms · Identity and authorityCost Engineering Agentic Workloads on AWS: Bedrock Inference and AgentCorePlatform economics · Cloud platforms · Agent architectureIdentity and Policy for Agents on AWS: AgentCore, Cedar, IAM, and Scoped AuthorityIdentity and authority · Cloud platforms · Agent architectureReading the AWS Well-Architected Agentic AI Lens in PracticeReliability and recovery · Production systems · Cloud platforms

Design

Interfaces and knowledge for human-agent teams
Designing Human-Agent Handoffs for Regulatory ScrutinyHuman oversight · Risk and regulation · Interface designDesigning Interfaces for Human-Agent TeamsInterface design · Human oversight · Reliability and recoveryEvidence-First UX: Showing Claims with Their ProvenanceEvidence and provenance · Interface design · Identity and authorityKnowledge Architecture as a Design DisciplineIdentity and authority · Production systems · Retrieval systemsThe Autonomy Ledger: Making Agent Permissions LegibleIdentity and authority · Interface design · Production systems

Research

Current research translated into design decisions
Agent Research 2023–2026: What Held up in ProductionProduction systems · Agent architecture · Reasoning and planningBefore the Agent Acts: Governing Context Acquisition under Cost, Privacy and ConsequenceEvidence and provenance · Identity and authority · Context engineeringModel-Risk Evidence for Agentic Systems: From Outcome Accuracy to Trajectory AssuranceEvaluation and assurance · Production systems · Risk and regulationTest-Time Compute in the Enterprise: When Reasoning Tokens are Worth MoneyPlatform economics · Reasoning and planning · Transformer mechanismsThe Reasoning Stack, 2026Reasoning and planning · Production systems · Reliability and recoveryWhat Scaling Laws Don't Tell YouTransformer mechanisms · Production systems · Learning systems

Machine Intelligence

First principles of representation, control and search
What Would Count as Intelligence?Research methods · Learning systems · Causality and interventionProblems, Tasks and EnvironmentsLearning systems · Reliability and recovery · Research methodsComputation as State TransformationCausality and intervention · Research library · Reliability and recoveryAlgorithms, Heuristics and Learned ProceduresReasoning and planning · Learning systems · Production systemsRepresentation Before IntelligenceRepresentation · Causality and intervention · Learning systemsFeedback, Control and the Cybernetic LoopReliability and recovery · Causality and intervention · Learning systemsLearning, Reasoning and Search are not SynonymsCausality and intervention · Reasoning and planning · Learning systemsFrom Programme to AgentProduction systems · Reasoning and planning · Agent architectureSearch as Structured ExplorationReasoning and planning · Platform economics · Identity and authorityHeuristics, a* and the Price of GuidanceReasoning and planning · Platform economics · Evidence and provenanceAdversarial Search and Strategic IntelligenceLearning systems · Security and containment · Reasoning and planningPlanning Before ActionReasoning and planning · Production systems · Reliability and recoveryKnowledge Representation and Logic: The Price of ExplicitnessRepresentation · Identity and authority · Causality and interventionReinforcement Learning and Sequential ChoiceLearning systems · Production systems · Causality and interventionThe Bitter Lesson and the Neuro-Symbolic ReplyLearning systems · Production systems · Reasoning and planningVectors as Directions, not ListsRepresentation · Production systems · Research methodsMatrices as MachinesRepresentation · Learning systems · Production systemsEigenstructure and Persistent DirectionsRepresentation · Production systems · Research methodsOptimisation, Approximation and GuaranteesLearning systems · Research methods · Platform economics

Consciousness

Machine consciousness research and assuranceOpen series map →
A Comparative Atlas of Possible MindsConsciousness science · Causality and intervention · Research methodsA Consciousness Assurance Case: Governing Machines We Cannot ClassifyConsciousness science · Causality and intervention · Evaluation and assuranceBefore Experience: The Minimal ConstraintsConsciousness science · Research methods · Causality and interventionBinding and the Candidate-Subject BoundaryConsciousness science · Causality and intervention · Identity and authorityBuilding an Attention Schema Without Mistaking Report for ExperienceConsciousness science · Causality and intervention · Transformer mechanismsCounting and Combinatorial ExplosionReasoning and planning · Platform economics · Research libraryDoes the Substrate Matter, or the Causal Organisation?Causality and intervention · Consciousness science · Research methodsEmergence and the Ruler that Creates ItConsciousness science · Causality and intervention · Production systemsFirst-Person, Interpersonal and Third-Person EvidenceResearch methods · Consciousness science · Causality and interventionFive Words People Confuse: Intelligence, Consciousness, Sentience, Selfhood and AgencyConsciousness science · Research methods · Causality and interventionFrom Observation to Explanation: Correlation, Causation and ConstitutionCausality and intervention · Research methods · Identity and authorityFunctions and Compositional Behaviour: Why the Model is the Path, not the PileTransformer mechanisms · Causality and intervention · Research methodsHow to Read a Frontier ClaimResearch methods · Causality and intervention · Evidence and provenanceIntegrated Information and the Causal Structure of Artificial SystemsCausality and intervention · Consciousness science · Production systemsLevels, Boundaries and the Grain of ExplanationCausality and intervention · Research methods · Reliability and recoveryNumber, Measurement and RepresentationRepresentation · Research methods · Risk and regulationResearch Without Theory VictoryResearch methods · Consciousness science · Evidence and provenanceRights, Forks and Interrupted MindsConsciousness science · Evidence and provenance · Production systemsSets, Logic and ProofReasoning and planning · Causality and intervention · Evidence and provenanceThe Hard Problem and the Price of Every ExitConsciousness science · Causality and intervention · Learning systemsThought Experiments as Instruments, not OraclesConsciousness science · Causality and intervention · Research methodsThree Ledgers: Function, Mechanism and ExperienceCausality and intervention · Consciousness science · Research methodsWhat Counts as Evidence for Another Mind?Causality and intervention · Evidence and provenance · Consciousness science

Search the complete catalogue →