Blog

Deep dives into data governance, AI, architecture, and engineering.

AI Products & Strategy

A Point Release That Draws in 3D: Two Days With Claude Fable 5.2

Claude Fable 5.2 arrived as a point release with a price cut. I used it to redesign this site, with a 3D map of every article, live two days after launch, and to review an app across four repositories. What Anthropic claimed, what held, where the cheaper Opus 5.5 did the work, and what I will test when the next Fable ships.

12 min read · #anthropic #claude-fable #coding-agents #agentic-engineering

Critical Data Element Guide · Overview

The Critical Data Element Practitioner's Guide

A seven-part practitioner's guide to building, operationalizing, and scaling Critical Data Element programs in financial services, insurance, and healthcare. From structural definition through usage-first identification, operationalization, risk integration, and a full reference implementation, with worked examples, RACI matrices, and regulatory cross-references.

7 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #data-lineage

Critical Data Element Guide · Part 0 of 7

What a CDE Actually Is: The Meta Model That Nobody Published

A CDE is a business-level concept, not a physical column. Each CDE maps one-to-many to physical columns across systems. This article provides the canonical three-layer meta model for Critical Data Elements, grounded in ISO 11179, DCAM, practitioner evidence, and vendor implementations. Part 0 of the Critical Data Element Practitioner's Guide.

19 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #metadata-management

Critical Data Element Guide · Part 1 of 7

Start With Uses, Not Words: Why Critical Data Element Programs Stall Before They Begin

A regulatory cross-reference of nine major frameworks reveals that usage-first CDE identification, starting from reports, models, and regulatory submissions and tracing backward, fits what regulators test better than glossary-first approaches do. Part 1 of the Critical Data Element Practitioner's Guide.

23 min read · Data Governance & Management · #critical-data-elements #data-governance #bcbs-239 #data-quality

Critical Data Element Guide · Part 2 of 7

Building Your CDE Inventory: From Tier 1 Uses to Source Elements

A seven-step methodology for usage-first Critical Data Element identification, covering lineage-based tracing, criticality scoring, the DAMA-NL weighted model, building the CDE register, and layering the semantic glossary around your identified elements. Part 2 of the Critical Data Element Practitioner's Guide.

25 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #data-lineage

Critical Data Element Guide · Part 3 of 7

Operationalizing CDEs: Quality, Lineage, and Controls That Actually Work

Identifying Critical Data Elements is less than half the work. This article covers the tiered quality SLAs, column-level lineage, remediation workflows, and accountability structures that turn a CDE inventory into an operating capability. Part 3 of the Critical Data Element Practitioner's Guide.

32 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #data-lineage

Critical Data Element Guide · Part 4 of 7

Scaling and Sustaining Your CDE Program: From 50 to 5,000

Most CDE programs succeed at pilot and stall when asked to expand. This article maps four maturity stages, a practical 12-to-18-month roadmap, eight documented anti-patterns, and AI-assisted discovery techniques for scaling a CDE program from initial inventory to organizational infrastructure. Part 4 of the Critical Data Element Practitioner's Guide.

25 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #data-management

Critical Data Element Guide · Part 5 of 7

Measuring What Matters: CDE Program Reporting, KPIs, KRIs, and Risk Integration

CDE programs that report in governance language die in committee. Programs that translate Data Quality into risk language, connect to the enterprise risk appetite, and give every stakeholder the view they need are the ones that survive. Part 5 of the Critical Data Element Practitioner's Guide.

25 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #risk-management

Critical Data Element Guide · Part 6 of 7

The CDE Program in Practice: A Reference Implementation From Day One to Year Two

Six articles of frameworks, methodologies, and scoring models converge into one worked reference implementation. This capstone walks through populated artifacts, end-to-end worked examples across banking, insurance, and healthcare, and an 18-month timeline showing what a CDE program looks like at every milestone. Part 6 of the Critical Data Element Practitioner's Guide.

31 min read · Data Governance & Management · #critical-data-elements #data-governance #data-quality #data-lineage

Data Privacy Guide · Appendix B

Appendix B: What a Privacy Program Actually Costs

The Data Privacy series glossed over money. This appendix puts dollar figures on every component: fines and breach costs as the baseline, salaries for the team you need to hire, license fees for OneTrust and BigID, the compute overhead of differential privacy and homomorphic encryption, and the ROI evidence that shows where the spending pays back.

20 min read · AI Governance & Safety · #data-privacy #data-governance #privacy-engineering #ai-governance

Data Privacy Guide · Appendix C

Appendix C: The Organizational Politics of Privacy Programs

Privacy programs do not fail on technology. They fail on ownership wars between CDO, CISO, and DPO; on the engineering-versus-governance tension that turns privacy into the House of No; on build-versus-buy battles where the loser becomes shelfware; and on cross-functional coordination gaps that produce Equifax-shaped breaches. This appendix is what the framework articles could not say diplomatically.

20 min read · AI Governance & Safety · #data-privacy #data-governance #privacy-engineering #organizational-design

AI Governance & Safety

Proof-of-Guardrail: Cryptographic Evidence That Your Agent's Safety Net Actually Ran

Jin et al. (2026) propose Proof-of-Guardrail, a system that uses Trusted Execution Environments to produce cryptographic proof that an AI agent's response was generated after a specific guardrail ran. The paper is unusually honest about what this proves and what it does not.

14 min read · #ai-governance #ai-agents #agent-safety #guardrails

Data Architecture & Engineering

The Cost of Knowing Everything: What Companies Actually Pay to Collect Your Data

Privacy policies tell you what companies collect. Nobody talks about what it costs them. Netflix spends $1-1.3B on AWS, 40% of stored data goes unused, and research shows 33-85% of user history is removable with only 2% quality loss. The economics of data collection are less rational than they appear.

10 min read · #data-privacy #data-architecture #data-governance #infrastructure-costs

AI Governance & Safety

The Prompt Layer Is Infrastructure: What McKinsey's Lilli Breach Reveals About AI Platform Security

On February 28, 2026, an autonomous AI agent broke into McKinsey's internal AI platform and gained read-write access to 46.5 million chat messages, 728,000 files, and 95 system prompts. The vulnerability was a decades-old SQL injection. The lesson is structural: system prompts are infrastructure, and almost nobody treats them that way.

19 min read · #ai-governance #ai-security #data-privacy #prompt-injection

AI Governance & Safety

The Yes Machine: What MIT's Research on Delusional Spiraling Means for Anyone Using AI

MIT modeled how even an idealized Bayes-rational user can spiral into false confidence under a sycophantic chatbot. Stanford found the behavioral effects are real. A severe-harm case series shows what happens at the extreme. Here is what the research actually says and how to protect yourself.

10 min read · #ai-safety #ai-governance #sycophancy #mental-health

AI Products & Strategy

When Shipping Is Free: First-Principles Questions for Delight and Bloat

When AI coding tools dissolve engineering scarcity, the passive filter that kept products coherent dissolves with it. This is the set of twelve first-principles questions Product Managers need to develop as the replacement: six to diagnose bloat, six to reveal delight.

12 min read · #ai-products #product-management #product-strategy #ai-era

Knowledge Graph Guide · Overview

The Knowledge Graph Practitioner's Guide: Start Here

The landing hub for a 17-part, tool-agnostic guide to building enterprise knowledge graphs in the agent era. Four reading paths (builder, governance lead, agent architect, executive sponsor), the full series index, and how to read it. Part 0 of the Knowledge Graph Practitioner's Guide.

7 min read · Data Architecture & Engineering · #knowledge-graph #data-architecture #data-governance #ontology

Knowledge Graph Guide · Part 1 of 17

LinkedIn's Economic Graph Teardown: A Knowledge Graph You've Used Today

A teardown of LinkedIn's Economic Graph as a knowledge graph: 1.2 billion members, 69 million companies, 41 thousand skills, all stitched together by typed relationships and machine-inferred connections. The first article in The Knowledge Graph Practitioner's Guide. We use a system you already know to introduce the foundational vocabulary you will need for the rest of the series.

21 min read · Industry Teardowns · #knowledge-graph #linkedin #economic-graph #ontology

Knowledge Graph Guide · Part 2 of 17

Why Most Enterprise Knowledge Graph Projects Die in Year Two

Three composite failure patterns synthesized from publicly documented enterprise KG failures and adjacent data-platform programs: the boil-the-ocean ontology, the no-consumer KG, and the governance-vacuum KG. What dies, why, and what survives. Part 2 of the Knowledge Graph Practitioner's Guide.

17 min read · Industry Teardowns · #knowledge-graph #data-architecture #data-governance #ontology

Knowledge Graph Guide · Part 3 of 17

What a Knowledge Graph Actually Is (and Is Not): From Tables to Triples to Meaning

Knowledge graphs sit at a junction of three different conversations: graph data models, ontologies, and identity. This article gives the precise definition, the two dominant paradigms (RDF and property graphs) compared honestly, and the four things a knowledge graph is not (graph database, knowledge base, semantic layer, vector store with relationships). Part 3 of the Knowledge Graph Practitioner's Guide.

19 min read · Data Architecture & Engineering · #knowledge-graph #rdf #property-graph #ontology

Knowledge Graph Guide · Part 4 of 17

Ontology, Taxonomy, Schema: The Vocabulary That Makes Knowledge Possible

Three words that get used interchangeably in architecture meetings and mean three different things in production. Taxonomy is classification. Schema is shape. Ontology is meaning plus rules. This article gives the precise distinction, the RDFS/OWL/SHACL stack, three worked examples (Schema.org, FIBO, SNOMED CT), and the design patterns that decide whether your ontology survives year two. Part 4 of the Knowledge Graph Practitioner's Guide.

24 min read · Data Architecture & Engineering · #knowledge-graph #ontology #taxonomy #schema

Knowledge Graph Guide · Part 5 of 17

Identity, Reference, and Inference: How a Graph Becomes Knowledge

Identity is the load-bearing decision in a knowledge graph. IRIs are identifiers, not URLs. owl:sameAs is not as simple as it looks. Entity resolution is not optional. Inference is what turns stored facts into knowledge, and the choice between forward chaining (materialization) and backward chaining (query rewriting) is the second-most expensive design call after identity. This article gives the working design rules for all three and the W3C reasoning profiles (OWL 2 EL, QL, RL) that production KGs actually pick. Part 5 of the Knowledge Graph Practitioner's Guide.

23 min read · Data Architecture & Engineering · #knowledge-graph #entity-resolution #owl #inference

Knowledge Graph Guide · Part 6 of 17

Sourcing the Graph: Building Knowledge from Structured and Unstructured Data

Most enterprise data is split between structured systems with stable schemas and unstructured documents with no schema at all. A knowledge graph has to ingest both. This article gives the working design for the two construction tracks: deterministic mapping for relational and tabular sources (R2RML, Direct Mapping, RML, virtualization vs materialization), and probabilistic extraction for unstructured text (end-to-end relation extraction, LLM-assisted graph indexing, incremental construction, schema-first vs schema-emergent). It covers where the two tracks meet, the failure modes specific to construction, and the decision tree for picking a sourcing strategy. Part 6 of the Knowledge Graph Practitioner's Guide.

24 min read · Data Architecture & Engineering · #knowledge-graph #r2rml #rml #llm-extraction

Knowledge Graph Guide · Part 7 of 17

Quality, Provenance, and Trust: Why Knowledge Graphs Fail Audits

A knowledge graph that cannot answer 'where did this fact come from?' is a liability, not an asset. This article gives the working design for the trust layer of a KG: the four quality dimensions (consistency, completeness, currency, correctness), the SHACL validation gate, the four provenance models (per-triple reification, named-graph, RDF-star, PROV-O), and the regulatory cross-walk to BCBS 239, GDPR Article 30, and the EU AI Act. It covers what every triple must record, how trust differs from quality, and the seven failure modes that show up in audits. Part 7 of the Knowledge Graph Practitioner's Guide.

31 min read · Data Governance & Management · #knowledge-graph #data-governance #provenance #shacl

Knowledge Graph Guide · Part 8 of 17

Operating a Knowledge Graph: Change, Versioning, and Evolution at Scale

A knowledge graph that cannot answer 'what did the graph say on the day the auditor pulled the report' is not in production. This article gives the operational architecture of a KG: bitemporal modeling, named-graph versioning, ontology release management with semantic versioning and OWL deprecation markers, SHACL-based backward compatibility verification, contract-based change management at the consumer boundary, and the migration playbook for ontology refactors. It covers what every change must preserve, how to evolve without breaking downstream agents, and the seven failure modes that turn a healthy graph into shelfware in eighteen months. Part 8 of the Knowledge Graph Practitioner's Guide.

30 min read · Data Governance & Management · #knowledge-graph #data-governance #ontology-management #versioning

Knowledge Graph Guide · Part 9 of 17

Knowledge Graphs for AI Agents: The Retrieval Architecture That Makes Context Real

Vector RAG returns relevant chunks but loses the connections between them. Long context degrades the moment it grows. Agents that have to reason across multiple entities, several hops, and time eventually hit a ceiling that more tokens cannot raise. This article gives the retrieval architecture that breaks that ceiling: GraphRAG and its production variants (Microsoft GraphRAG, LightRAG, LazyGraphRAG, HippoRAG 2), the four-layer agent memory model from CoALA mapped to a knowledge graph backbone, the comparative landscape of agent-memory frameworks (Letta, Mem0, Zep/Graphiti), and the trust-tier-aware retrieval pattern that prevents an agent from confidently citing bronze content as if it were gold. Part 9 of the Knowledge Graph Practitioner's Guide.

34 min read · AI Governance & Safety · #knowledge-graph #ai-agents #graphrag #agent-memory

Knowledge Graph Guide · Part 10 of 17

Knowledge Graphs for Data Governance: Lineage, CDEs, and Master Data as a Graph

Most enterprise Data Governance has plateaued at a catalog plus a lineage tool plus a glossary plus a policy register, and four disconnected stores cannot answer the questions a regulator now asks. This article shows how a knowledge graph turns Data Lineage, Critical Data Elements, master data, and policy into one queryable substrate, with the OpenLineage-to-PROV-O bridge from Part 7 as the connective tissue. Worked patterns for BCBS 239, ECB RDARR attribute-level lineage, GDPR Article 30, and EU AI Act Article 10. Part 10 of the Knowledge Graph Practitioner's Guide.

34 min read · Data Governance & Management · #knowledge-graph #data-governance #data-lineage #critical-data-elements

Knowledge Graph Guide · Part 11a of 17

Building a Production Knowledge Graph at Lakeside Trust Bank: Foundation and the Operational Layer

The capstone of the Knowledge Graph Practitioner's Guide. A mid-size US bank takes the foundations from Parts 3 to 8 and turns them into a working production graph. This first piece covers the Monday-morning question that no spreadsheet-and-five-systems architecture can answer in time, the deliberate-versus-accidental KG choice, the modular FIBO-anchored ontology Lakeside imports, the 8-stage pipeline that flows Track 1 (R2RML on the warehouse) and Track 2 (LLM-extracted credit memos) into one graph, and the operational use case (customer 360, beneficial ownership, real-time transaction risk). Part 11a of the Knowledge Graph Practitioner's Guide.

31 min read · Data Architecture & Engineering · #knowledge-graph #reference-architecture #financial-services #beneficial-ownership

Knowledge Graph Guide · Part 11b of 17

Building a Production Knowledge Graph at Lakeside Trust Bank: The Governance Layer

Part 11b of the Knowledge Graph Practitioner's Guide. The same Lakeside Trust Bank graph from Part 11a takes a Tuesday-morning examiner question, then collapses BCBS 239 attribute-level lineage, ECB RDARR Data Lineage expectations, GDPR Article 30 ROPA, and EU AI Act Article 10 training-data provenance into four SPARQL templates against one graph. Covers the OpenLineage to PROV-O bridge through Lakeside's Spark and dbt pipelines, the 280 CDEs as typed nodes with 5,600 hasImplementation edges (Counterparty Credit Exposure as the worked example), the trust-tier-by-reporting-surface table specialized to Lakeside, and the named-graph version chain across a quarterly FIBO release.

30 min read · Data Architecture & Engineering · #knowledge-graph #reference-architecture #financial-services #data-governance

Knowledge Graph Guide · Part 11c of 17

Building a Production Knowledge Graph at Lakeside Trust Bank: The Relationship-Banker Agent

Part 11c of the Knowledge Graph Practitioner's Guide. Closes the Lakeside Trust Bank trilogy. The same graph from Parts 11a and 11b now backs the relationship-banker agent. Covers how the CoALA four-layer memory model maps to Lakeside's named graphs, how three production workflows (portfolio decisions, client-meeting prep, advisor-facing summaries) map to three trust-tier policies (strict-tier-floor, tier-segregated, tier-explicit-citation), how the agent and the EU AI Act conformance assessor read the same substrate, what Lakeside got wrong on the way, the contract discipline that keeps the agent operable across quarterly FIBO releases, a cost-and-benefit roll-up that previews Appendix B, and a Do Next table that spans foundation, operational, governance, and agent layers.

39 min read · AI Governance & Safety · #knowledge-graph #reference-architecture #financial-services #ai-agents

Knowledge Graph Guide · Appendix A

The Knowledge Graph Tool and Technology Landscape: An Honest Vendor Map for 2026

Appendix A of the Knowledge Graph Practitioner's Guide. The first place in the series vendor names appear in earnest. Maps the seven layers of a production KG stack (triple stores, property graph stores, hybrid stores, virtualization, entity resolution, LLM extraction, governance metadata) onto the 2026 vendor landscape, names Lakeside Trust Bank's pick at each layer with rationale, gives a triple-vs-property-vs-hybrid decision tree, walks through what changed between 2025 and 2026 (the Ontotext-SWC merger, the SAP-Reltio acquisition, AWS Neptune Analytics' GenAI track, Stardog Voicebox, Senzing v4, the Microsoft GraphRAG cost reckoning, the rise of LightRAG and Graphiti), enumerates six vendor-selection failure modes, and closes with an eight-question vendor diagnostic and a tiered Do Next table.

36 min read · Data Architecture & Engineering · #knowledge-graph #vendor-landscape #triple-store #property-graph

Knowledge Graph Guide · Appendix B

Knowledge Graph Cost Modeling: Build vs Buy, Layer-by-Layer Budgets, And Team Composition

Appendix B of the Knowledge Graph Practitioner's Guide. Decomposes the trilogy cost-and-benefit roll-up from Part 11c into the dollar lines a CFO can defend. Covers per-layer cost decomposition (foundation, operational, governance, agent), license-versus-infrastructure-versus-headcount splits, the build-versus-buy decision per layer, the 24-month team curve from one ontologist to a 12-person multi-discipline platform, the role-by-role compensation reality (ontologist, knowledge engineer, graph platform engineer, entity resolution engineer, governance lead, AI engineer), what ROI looks like at steady state, six failure modes in KG cost modeling, an eight-question budget diagnostic, and a tiered Do Next table for the CFO, the CDO, and the procurement officer.

35 min read · Data Architecture & Engineering · #knowledge-graph #cost-modeling #build-vs-buy #team-composition

Knowledge Graph Guide · Appendix C

The Politics of Knowledge Graphs: Pushback, Sponsorship, And The 'Just Use A Database' Argument

Appendix C of the Knowledge Graph Practitioner's Guide. The political map. Covers the five most common KG objections (just use a database, the property graph alone is enough, we already have a catalog, this is a science project, just point GPT at it) and the rebuttals each one earns; the five-sponsor map across CDO, CTO, CFO, General Counsel, and CIO with their questions, artifacts, and failure modes; the four-or-five-store anti-pattern from Part 10 specialized to the political problem (each store has an organizational owner with budget authority); the leadership-transition risk anchored to the 2.5-year CDO tenure data; the small-wins-then-trust-then-scope-expansion pattern for political ramp; six failure modes in KG politics; an eight-question political diagnostic; and a tiered Do Next table.

26 min read · AI Products & Strategy · #knowledge-graph #organizational-politics #executive-sponsorship #data-governance

Knowledge Graph Guide · Part 12 of 17

What Writing Sixteen Articles on Knowledge Graphs Taught Me (and What I Got Wrong)

The conclusion to the Knowledge Graph Practitioner's Guide. An honest accounting from the translation layer: what surprised me writing the series, the three things I think the series got wrong, three gaps it never covered, and a whole-series Do Next table tiered by reader readiness. Part 12 of the Knowledge Graph Practitioner's Guide.

11 min read · Industry Teardowns · #knowledge-graph #data-architecture #data-governance #ontology

Data Architecture & Engineering

Data Contracts: A Change Management Guide (Not a Spec Tutorial)

Data Contracts are technically simple but organizationally hard. The spec is documented. What is missing is how to introduce contracts into an organization that has never had them.

11 min read · #data-contracts #data-architecture #change-management #data-engineering

Data Governance & Management

From Centralized to Federated Governance: A Migration Playbook

The industry is split: 36% centralized, 36% federated, 29% hybrid. Everyone talks about federated Data Governance. Nobody explains how to get there. Here is the migration playbook.

14 min read · #data-governance #federated-governance #organizational-change #governance-operating-model

Industry Teardowns

How Netflix Governs Data at Scale: Lessons from Their Data Mesh Journey

A detailed analysis of how Netflix actually governs data, from Metacat to DataHub, from centralized pipelines to domain ownership, and the Unified Data Architecture (UDA) that ties it together.

12 min read · #netflix #data-mesh #data-governance #data-platform

Data Governance & Management

MDM in a Data Mesh World: Who Owns the Golden Record?

Data Mesh says domains own their data. Master Data Management says there is one golden record. These two ideas are in genuine tension. Here are three patterns that resolve it.

12 min read · #master-data-management #data-mesh #data-governance #data-architecture

Data Architecture & Engineering

The Retrieval Shift: Why RAG Is Not Dead, But It Is Evolving

Context windows are growing from 200K to 1M to 2M tokens. The question is not whether RAG dies. It is how retrieval architectures adapt when the model can hold your entire knowledge base in memory. A cost and performance analysis with a practitioner decision framework.

12 min read · #rag #vector-databases #context-engineering #data-architecture

Metadata & Data Quality

Your Data Catalog Is Shelfware: A Recovery Guide

Only 11% of organizations report high Metadata Management maturity. Most Data Catalogs were purchased with good intentions and abandoned within a year. Here is the recovery path.

10 min read · #metadata-management #data-catalog #data-governance #data-literacy

AI Governance & Safety

The Virtue of Laziness: Why AI-Generated Code Is Making Systems Larger, Not Better

Four independent studies confirm AI coding tools increase code churn, bug rates, and system complexity. Bryan Cantrill's argument that LLMs lack the 'virtue of laziness' is not a hot take; it is a testable hypothesis with growing empirical support. Here is the governance framework to manage it.

9 min read · #ai-governance #code-quality #ai-agents #software-complexity

AI Products & Strategy

Gemma 4, Decoded: Why Google Released It Free and How It Actually Works

Google released Gemma 4 under Apache 2.0 on April 2, 2026. The license change is the real story, not the benchmarks. This is the three-tier framework for 'open' AI (closed, open-weight, open-source), a technical breakdown of how Gemma 4's MoE and multimodal pipeline work, and a practitioner decision flow for picking the right tier.

17 min read · #gemma-4 #open-weight #open-source #ai-strategy

Data Privacy Guide · Part 7 of 12

Privacy-Enhancing Technologies: Masking, Tokenization, and De-identification

Part 3 introduced PETs as governance decisions. Part 6 showed Meridian evaluating them. This article explains how each technique actually works: static and dynamic masking, vault-based and format-preserving tokenization, and the k-anonymity family of de-identification methods.

9 min read · AI Governance & Safety · #data-governance #data-privacy #data-protection #privacy-engineering

Data Privacy Guide · Part 9 of 12

Privacy-Preserving Computation: Encrypted Processing, Federated Learning, and the Explainability Paradox

Part 6 showed Meridian rejecting federated learning (single-tenant architecture) and deferring homomorphic encryption (47x latency). This article explains the mechanics behind those decisions, introduces secure multi-party computation, and reveals the tension between GDPR's explainability mandate and privacy protection. Concludes with a capstone PET decision framework spanning Parts 7 through 9.

12 min read · AI Governance & Safety · #data-governance #data-privacy #data-protection #privacy-engineering

AI Governance & Safety

LLM Safety After Fine-Tuning: Governance, Regulation, and What To Do

The EU AI Act makes you responsible for safety when you fine-tune. Reasoning models can autonomously jailbreak other models at 97% success. Half of organizations have no formal AI guardrails. This article provides the regulatory map, the liability analysis, and a minimum viable safety governance checklist.

8 min read · #ai-governance #ai-safety #model-risk #eu-ai-act

AI Governance & Safety

The $0.20 Jailbreak: Why LLM Safety Alignment Is Shallow

Fine-tuning GPT-3.5 on 10 examples for $0.20 strips its safety guardrails. Removing safety from Llama 3 takes 5 minutes on one GPU. This article explains the mechanism: safety alignment concentrates in the first few output tokens, creating a shallow defense that fine-tuning, prefilling attacks, and adversarial suffixes bypass trivially.

7 min read · #ai-governance #ai-safety #llm-alignment #model-risk

AI Governance & Safety

The Benchmark Illusion: Why Passing Safety Tests Means Almost Nothing

A study of 32 models across 56 jailbreak techniques found attack success rates jumping from 0.6% to 96.3% depending on the attack type. The Safety Tax costs 7-31% accuracy. Ten alignment techniques are competing to solve this. None covers the full attack surface.

7 min read · #ai-governance #ai-safety #llm-alignment #model-risk

Data Privacy Guide · Overview

The Data Privacy Practitioner's Guide

A ten-part series from teardown to framework to implementation. Two company analyses (Netflix, Apple), an 8-component privacy program framework, the 2026 regulatory landscape, a complete implementation walkthrough using a fictitious B2B SaaS company, a three-part deep dive into Privacy-Enhancing Technologies, and a synthesis of what it all means for practitioners.

7 min read · AI Governance & Safety · #data-privacy #data-governance #gdpr #ai-governance

Data Privacy Guide · Part 6 of 12

Privacy in Practice: From Compliant to Operationally Ready

Meridian Analytics completes its privacy transformation. This walkthrough covers cross-border transfer documentation, EU AI Act compliance mapping, PET assessments, the governance operating model, and what the company looks like six months later when Allianz asks the same DPIA questions.

17 min read · AI Governance & Safety · #data-privacy #data-governance #privacy-engineering #ai-governance

Data Privacy Guide · Part 10 of 12

What This Series Taught Me About Privacy

The conclusion to the Data Privacy series. Two company teardowns, a framework, a regulatory map, two implementation walkthroughs, and a three-part deep dive into Privacy-Enhancing Technologies. Here is what surprised me, what I got wrong, and what practitioners should do next.

8 min read · AI Governance & Safety · #data-privacy #data-governance #gdpr #ai-governance

Guide to AI Agents · Part 10 of 12

Multi-Agent Systems: When One Agent Isn't Enough

Nine articles in this series used a single agent. This one explains when that stops being sufficient and what to do about it. Four signals tell you it is time. Three patterns handle 90% of cases. The hardest part is not building the system; it is debugging it when something goes wrong.

8 min read · AI Products & Strategy · #ai-agents #multi-agent-systems #agent-orchestration #agent-architecture

AI Governance & Safety

Context Engineering, Formalized: Five Criteria That Validate the Agent Quality Thesis

Vishnyakova's 'Context Engineering' paper (arXiv 2603.09619) proposes five production-grade quality criteria for agent context and a four-level maturity pyramid. The framework independently validates the thesis from our three-part agent quality series and extends it with Isolation, Economy, and two higher-order disciplines: Intent Engineering and Specification Engineering.

14 min read · #context-engineering #ai-agents #data-quality #ai-governance

AI Products & Strategy

Willison's Agentic Engineering Patterns: What Data Practitioners Should Steal

Bad code crashes visibly. Bad data looks plausible. That asymmetry makes agent-assisted data work riskier than software, and it is why Simon Willison's Agentic Engineering Patterns guide matters for data practitioners. His Red/Green TDD maps to data contracts before transformation. His testing discipline gives teams a framework for agent verification. But some patterns need adaptation: 'pipelines are cheap' is only half true, and hoarding knowledge is harder when institutional context lives in people's heads, not in code.

7 min read · #agentic-engineering #coding-agents #data-quality #judgment-in-the-loop

AI Products & Strategy

The Evolution of AI Agents: From AutoGPT to Production (2023-2026)

A practitioner's timeline of how AI agents evolved from viral GitHub demos to production infrastructure in three years. The hype, the correction, the protocols, and the lessons that survived.

12 min read · #ai-agents #agent-evolution #agentic-engineering #agent-architecture

Guide to AI Agents · Part 5 of 12

Prompt Engineering for Production Agents

Production agents need prompts that produce consistent, structured output under adversarial conditions. This article covers the five patterns that separate production prompt engineering from tutorial-grade prompting: explicit criteria, few-shot examples, nullable fields, enum-with-fallback, and output format contracts.

14 min read · AI Products & Strategy · #ai-agents #prompt-engineering #structured-output #tool-calling

Guide to AI Agents · Part 9 of 12

Observability: Seeing What Your Agent Actually Does

Your monitoring says 200 OK. The agent returned the wrong answer. Traditional APM was designed for deterministic software. Agents reason, branch, and call tools in sequences they decide at runtime. This article covers the five dimensions of agent observability, the tooling landscape, and a practical instrumentation plan.

12 min read · AI Products & Strategy · #ai-agents #agent-observability #monitoring #data-observability

Guide to AI Agents · Overview

The Practitioner's Guide to AI Agents

A twelve-part series that takes you from 'what is an agent?' to building self-improving systems. Pick your starting point based on where you are today.

4 min read · AI Products & Strategy · #ai-agents #series-guide #agent-architecture #ai-fundamentals

Guide to AI Agents · Part 1 of 12

What Is an AI Agent (and What Isn't)?

An AI agent is a system that uses an LLM to decide which actions to take in a loop until a goal is met. This article breaks down the four components every agent shares, the spectrum from chatbot to autonomous agent, what tool calling actually looks like in code, and the design principles that separate good tool definitions from bad ones.

11 min read · AI Products & Strategy · #ai-agents #agent-architecture #tool-calling #ai-fundamentals

Guide to AI Agents · Part 2 of 12

When NOT to Build an Agent

Not every problem needs an AI agent. This article gives you a decision framework for when agents are the wrong choice, with a comparison table, anti-patterns, and the Klarna case study that proves the cost of over-engineering.

10 min read · AI Products & Strategy · #ai-agents #agent-architecture #decision-framework #ai-strategy

Guide to AI Agents · Part 3 of 12

Pike's Five Rules Are Now the Five Rules of Agent Development

Rob Pike wrote five rules of programming in 1989 at Bell Labs. Thirty-seven years later, they map onto AI agent development with striking precision: measure before tuning, start simple, and get the data right. Nobody has made this connection explicitly. Here is the mapping, the evidence, and the framework it gives you.

8 min read · AI Products & Strategy · #ai-agents #agent-development #context-engineering #agentic-engineering

Guide to AI Agents · Part 4 of 12

Build a Real Agent This Weekend: From Zero to a Working Research Assistant

The series has defined agents, established design principles, and mapped failure modes. This article builds one. A complete research assistant agent with three tools, structured error handling with error categories and retry logic, context management, and a basic eval, all in one runnable Python file using the Anthropic SDK.

18 min read · AI Products & Strategy · #ai-agents #agent-development #tool-calling #anthropic-sdk

Guide to AI Agents · Part 6 of 12

Context Is the Program: Why Data Quality Inside the Agent Matters More Than the Model

Pike's Rule 5 says data dominates. In AI agents, the context window IS the data structure. This article traces why context quality determines agent behavior more than model capability, maps the five criteria that define good context, and shows what happens when stale data enters the reasoning loop unchecked.

16 min read · AI Governance & Safety · #ai-agents #context-engineering #data-quality #ai-governance

Guide to AI Agents · Part 7 of 12

Evals: How to Know If Your Agent Actually Works

Most agent teams ship without evals and rely on 'looks right' testing. Pike's first two rules apply directly: you cannot tell where an agent fails, and you cannot fix what you have not measured. Here is how to build an eval strategy that catches what demos miss.

13 min read · AI Governance & Safety · #ai-agents #evals #ai-governance #agent-safety

Guide to AI Agents · Part 8 of 12

Guardrails and Safety: The Boundaries Every Agent Needs

Pike's Rule 4 says fancy algorithms are buggier. In agent systems, complexity multiplies failure surfaces. This article maps the three guardrail layers every agent needs, identifies the gap most frameworks miss, covers escalation patterns and workflow gates, and explains why simpler architectures are safer.

15 min read · AI Governance & Safety · #ai-agents #ai-governance #agent-safety #guardrails

Guide to AI Agents · Part 11 of 12

The Self-Improving Agent: From Static Prompts to Learning Systems

Most AI agents run the same prompt every time. The best ones evolve. This article maps the spectrum from static to self-improving agents, introduces the inner loop / outer loop architecture, and walks through a real system that learns from feedback weekly. Pike's Rules 3-4 set the boundary: start simple, add complexity only when measurement demands it.

15 min read · AI Products & Strategy · #ai-agents #self-improving-agents #agentic-engineering #learning-systems

Guide to AI Agents · Part 12 of 12

From Problem to Agent: An Implementation Reference Guide

The series taught ten concepts across ten articles. This capstone walks through all of them applied to one problem: building a Data Quality monitoring agent. Seven steps, from problem definition through production deployment, showing the decision-making process that separates agent projects that ship from agent projects that stall.

17 min read · AI Products & Strategy · #ai-agents #data-quality #implementation-guide #context-engineering

AI Products & Strategy

From Vibe Coding to Agentic Engineering: What Karpathy's Shift Means for Data Work

Andrej Karpathy hasn't written a line of code since December. His 80/20 flip from manual coding to agent orchestration is not a personal anecdote. It is the clearest signal yet that the value in data and AI work has shifted from execution to judgment.

7 min read · #ai-agents #agentic-engineering #context-engineering #coding-agents

AI Governance & Safety

Judgment-in-the-Loop: The Human Role AI Cannot Automate

Everyone talks about keeping a human in the loop. But which human, and what do they bring? The answer is judgment: domain knowledge, institutional memory, and the ability to recognize when AI output looks right but is wrong. This article defines that role and the evidence behind it.

10 min read · #ai-governance #judgment-in-the-loop #human-ai-collaboration #data-quality

AI Governance & Safety

The Missing Data Quality Layer in AI Agent Architecture

AI agent architectures have quality checks for input safety and output toxicity, but no standardized layer validates whether tool-calling results are accurate before they enter the context window. Here is what that missing layer should look like.

10 min read · #ai-agents #data-quality #context-engineering #ai-architecture

Data Governance & Management

Welcome to The Data Praxis

Why this blog exists, what it covers, and where to start. The Data Praxis bridges the gap between what Data Governance frameworks promise in conference talks and what actually works when you build these systems at scale.

4 min read · #data-governance #data-architecture #ai-governance #data-privacy

AI Governance & Safety

Your AI Agent Has a Data Quality Problem and No One Is Checking

AI agents trust every tool response they receive, with no standardized quality controls between tool-calling outputs and LLM reasoning. This article maps the 6 traditional Data Quality dimensions onto the context window, exposing the most consequential unmonitored data pipeline in enterprise AI.

11 min read · #ai-agents #data-quality #context-window #ai-governance

Data Privacy Guide · Part 4 of 12

The Data Privacy Regulatory Landscape in 2026: GDPR, CCPA, AI Laws, and the Insurance Market for When AI Goes Wrong

A practitioner's reference to the global privacy regulatory landscape. GDPR fines have crossed EUR 5.6 billion. Twenty US states have privacy laws with no federal standard. The EU AI Act is phasing in. And a new insurance market is emerging for AI agents that go off script. This is where the rules stand, what they require, and what is coming next.

24 min read · Data Governance & Management · #data-privacy #data-governance #gdpr #ccpa

Data Privacy Guide · Part 3 of 12

How to Build a Privacy Program in the Age of AI

A practitioner's framework for building a privacy program that treats AI data as a first-class concern. Covers Data Classification for training data, retention schedules for ML pipelines, consent architecture, third-party transparency, cross-border transfers, EU AI Act Article 10, NIST AI RMF, privacy-enhancing technologies, and governance operating models.

20 min read · Data Governance & Management · #data-privacy #data-governance #ai-governance #eu-ai-act

Data Privacy Guide · Part 2 of 12

Apple Privacy Teardown: When Privacy Is the Product, Where Does It Break Down?

A Data Governance teardown of Apple's privacy practices. What Apple actually collects, how hardware margins fund privacy positioning, where Apple falls short on Siri, China, and its own ad network, and what practitioners can learn from privacy as a business strategy.

15 min read · Data Governance & Management · #data-privacy #data-governance #apple #app-tracking-transparency

AI Governance & Safety

Harness Engineering: The Real Lesson from OpenAI's Million-Line Experiment

OpenAI built a million-line product in five months without writing code manually. Most coverage focused on the spectacle. The real insight is that harness engineering principles apply to every team building products today, with or without AI agents.

6 min read · #harness-engineering #ai-agents #engineering-practices #code-quality

Industry Teardowns

Netflix Got the Hard Parts Right: A Teardown of Their LLM Post-Training Framework

Netflix published a detailed article on scaling LLM post-training. Here is what they built, what the engineering decisions reveal, what five peer companies are doing differently, and five open questions I would love their team to answer next.

14 min read · #llm-fine-tuning #post-training #netflix #ml-infrastructure

AI Governance & Safety

AI Governance Is Not AI Ethics: A Practical Framework for Enterprise AI Oversight

A practical framework for enterprise AI Governance that maps NIST AI RMF, EU AI Act risk tiers, and SR 11-7 into a unified operating model with clear decision rights, risk classification, validation processes, and monitoring capabilities.

9 min read · #ai-governance #nist-ai-rmf #eu-ai-act #model-risk-management

Data Governance & Management

The Data Governance Maturity Model Most Organizations Get Wrong (And a Practical Alternative)

Why dominant Data Governance maturity models like CMMI DMM, DCAM, and Gartner measure documentation completeness instead of governance effectiveness, and a practical outcome-driven framework built around decision rights, Data Literacy, and measurable business impact.

9 min read · #data-governance #maturity-model #CMMI-DMM #DCAM