Author

Data Bricklaying Diary

Records practices, thoughts, and pitfalls on the data grunt-work journey, sharing content on data platforms, data analysis, data processing, data governance, knowledge graphs, and more. Less theory, more hands‑on, making complex data technologies simple.

Articles

Likes

Views

Comments

Latest from Data Bricklaying Diary

81 recent articles

Data Bricklaying Diary

Sep 15, 2026 · Backend Development

Beyond 'Everything Is a Plugin': Defining Production Responsibility in Agent Harness

The article argues that a composable agent runtime alone is insufficient for production; it requires action contracts for external side effects, version binding with impact analysis, and independent audit evidence for task completion, with clear responsibility boundaries across runtime, plugins, action gateways, and human operators.

Action ContractsAgent Runtimeaudit evidence

0 likes · 18 min read

Beyond 'Everything Is a Plugin': Defining Production Responsibility in Agent Harness

Data Bricklaying Diary

Sep 15, 2026 · R&D Management

Evaluating Ontology Intelligence Projects: 6 Dimensions Beyond Concept Counts

This article argues that ontology intelligence projects must be evaluated on end-to-end business outcomes rather than concept or triple counts, proposing a six-dimension framework—business value, semantic quality, decision reliability, action closure, security governance, and unit economics—with mandatory safety thresholds and phase-gated decisions to expand, optimize, restrict, or stop projects.

KnowledgeOpsOntology Intelligenceaction reliability

0 likes · 37 min read

Evaluating Ontology Intelligence Projects: 6 Dimensions Beyond Concept Counts

Data Bricklaying Diary

Sep 14, 2026 · R&D Management

Tech Selection: Derive from Constraints, Not Feature Rankings

The article argues that technology selection should be driven by architectural constraints and long-term operational responsibility rather than feature comparisons, emphasizing total lifecycle costs, standardized Golden Paths to reduce duplicate effort, and requiring verifiable evidence for any exceptions.

architectural constraintsarchitectural decision recordsgolden path

0 likes · 23 min read

Tech Selection: Derive from Constraints, Not Feature Rankings

Data Bricklaying Diary

Sep 14, 2026 · Artificial Intelligence

Can LLMs Auto-Build Ontologies? AI Finds Candidates, Business Defines Reality

This article explains how large language models accelerate ontology engineering through candidate discovery, formalization, and validation, but cannot replace human responsibility for defining business objects, rules, and risk boundaries, proposing a seven-step human-AI collaboration process with three-zone isolation to prevent AI from polluting production baselines.

GovernanceHuman-AI CollaborationLarge Language Models

0 likes · 25 min read

Can LLMs Auto-Build Ontologies? AI Finds Candidates, Business Defines Reality

Data Bricklaying Diary

Sep 13, 2026 · Artificial Intelligence

Agent Consensus ≠ Execution: Designing the Negotiation-to-Action Request Pipeline

The article argues that multi-agent consensus via dynamic negotiation (Liquid Interface) only produces candidate plans, not executable actions, and details how negotiation records must be versioned, policy-checked, authorized, and transformed into verifiable action requests with state validation and audit trails.

Liquid Interfaceaction requestsaudit trails

0 likes · 17 min read

Agent Consensus ≠ Execution: Designing the Negotiation-to-Action Request Pipeline

Data Bricklaying Diary

Sep 11, 2026 · Backend Development

Distributed Systems' Hardest Challenge: State Over Interfaces — Consistency, Idempotency & Async Tasks

This article explores why cross-service state management is harder than API design in distributed systems, using a refund example to illustrate how local transactions, idempotency, state machines, Outbox patterns, and manual intervention achieve recoverable convergence when operations span multiple services and external channels.

Idempotencyasync tasksconsistency

0 likes · 21 min read

Distributed Systems' Hardest Challenge: State Over Interfaces — Consistency, Idempotency & Async Tasks

Data Bricklaying Diary

Sep 10, 2026 · R&D Management

Agent Runs Tests ≠ Trustworthy Results: Building Independent Verification in AI-Native Development

The article explains why AI agents running tests doesn't ensure reliable verification, proposing a five-layer system — self-verification, independent verifier, test role, CI toolchain, and business acceptance — with version-linked evidence, protected acceptance criteria, design cross-checks, and clear human judgment boundaries for semantic, risk, and authorization decisions.

AI-Native DevelopmentCI/CDacceptance criteria

0 likes · 13 min read

Agent Runs Tests ≠ Trustworthy Results: Building Independent Verification in AI-Native Development

Data Bricklaying Diary

Sep 10, 2026 · Backend Development

Terminology Normalization Beyond Similarity: Coverage, Routing, Drift & Regression Testing

This article explains why terminology normalization requires a continuous governance framework covering candidate recall, risk-based routing, manual review, drift attribution, and regression testing — not just similarity scores — and details metrics, routing rules, drift types, acceptance criteria, and a minimum viable loop for production systems.

Data Governancecandidate recalldrift detection

0 likes · 18 min read

Terminology Normalization Beyond Similarity: Coverage, Routing, Drift & Regression Testing

Data Bricklaying Diary

Sep 9, 2026 · Artificial Intelligence

Three Graphs, Three Jobs: Loop, Task Graph & Plugin Runtime in Agent Systems

This article argues that complex Agent systems must separate three distinct structures: Loop handles node-level convergence, Task Graph manages work dependencies and coordination, and Plugin Runtime binds capabilities like models and tools; mixing them into a monolithic Agent leads to invisible boundaries and unrecoverable failures.

Agent ArchitectureComponent GraphContext

0 likes · 17 min read

Three Graphs, Three Jobs: Loop, Task Graph & Plugin Runtime in Agent Systems

Data Bricklaying Diary

Sep 9, 2026 · R&D Management

FDE: Converting Field Evidence into Runnable Business Models

This article explains how FDE creates a business source model from field evidence, defining roles, a five-step convergence process, traceability to engineering artifacts, and baseline release criteria, ensuring business semantics are confirmed, conflicts preserved, and models constrain implementation without replacing source systems.

FDEbusiness modelingbusiness source model

0 likes · 24 min read

FDE: Converting Field Evidence into Runnable Business Models