Agentic AI in production · 5 min read · with a concept loop
Writing.
Notes on funding, governing and operating AI in production.
Start here.
Strategy and funding
Economics
Governance
Agent reliability
Security
Architecture
Evaluation
The one story told from the inside: how Snowboard replaced an eight-figure automation contract.
The archive.
Keywords
Essays match every chosen keyword
86 essays
Evaluation and assurance · 5 min read · with a concept loop
Verification Engineering: How to Trust an AI System's Answer
Evaluation and assurance · 5 min read · with a concept loop
Why Confident AI Answers Still Need a Calibration Layer
Model architecture and training · 5 min read · with a concept loop
The Time-to-Quality Ledger for Training System Spending
Agentic AI in production · 5 min read · with a concept loop
The Four Gates Between a Tool Call and a Business Outcome
Model architecture and training · 5 min read · with a concept loop
Tokenization and Embedding Choices That Decide AI Cost and Risk
Strategy, product and value · 5 min read · with a concept loop
The Mechanism Ownership Map for AI Engineering Labels
Data, adaptation and alignment · 4 min read · with a concept loop
Synthetic Data and Curriculum Design: The Three-Gate Pipeline
Context, retrieval and memory · 5 min read · with a concept loop
Structured Output and the Validity Ladder for AI
Agentic AI in production · 5 min read · with a concept loop
The Sealed-Evidence Arbiter for Multi-Agent AI Risk
Model architecture and training · 5 min read · with a concept loop
State-Space and Recurrent Models: Audit What They Really Keep
Agentic AI in production · 5 min read · with a concept loop
Skill Engineering: The Three-Gate Contract for AI Agents
Context, retrieval and memory · 5 min read · with a concept loop
Why Semantic Caching Needs Authorization, Not Just Similarity
Agentic AI in production · 5 min read · with a concept loop
How to Govern Self-Improving AI Agents Without Losing Control
Secure LLM and agent execution · 5 min read · with a concept loop
The Four-Gate Boundary: Stopping AI Supply-Chain Attacks
Secure LLM and agent execution · 5 min read · with a concept loop
The Reconciliation Gate: How to Resume AI Agents Safely
Responsible AI and governance · 5 min read · with a concept loop
The Residual-Risk Chain: How to Prove an AI System Is Safe
Operations and economics · 5 min read · with a concept loop
AI Model Routing: How to Cut Cost Without Losing Quality
Context, retrieval and memory · 5 min read · with a concept loop
Retrieval and RAG: How AI Systems Ground Answers in Evidence
Strategy, product and value · 4 min read · with a concept loop
Requirements and Contracts: The Two-Gate Model for AI
Operations and economics · 5 min read · with a concept loop
Reliability Engineering for AI Agents: Contain Before You Recover
Context, retrieval and memory · 5 min read · with a concept loop
What Automatic Prompt Optimization Is Actually Worth
Strategy, product and value · 4 min read · with a concept loop
How to Judge an AI Product by Its Outcome, Not Its Output
Secure LLM and agent execution · 4 min read · with a concept loop
The Four-Boundary Check: Securing AI Access Across Tenants
Model architecture and training · 4 min read · with a concept loop
The Four-Layer Test Before Betting on Joint-Embedding Models
Agentic AI in production · 5 min read · with a concept loop
The Four-Bottleneck Model for AI Reasoning and Planning
Evaluation and assurance · 5 min read · with a concept loop
The History Ledger: How to Evaluate AI Agents That Adapt
Data, adaptation and alignment · 5 min read · with a concept loop
Open-Ended Curricula: Why More Environments Can Hurt Training
Context, retrieval and memory · 5 min read · with a concept loop
Knowledge Graphs and Ontologies: What They Are Worth Funding
Evaluation and assurance · 5 min read · with a concept loop
Why Most AI Failure Explanations Are Not Proof of Cause
Model architecture and training · 5 min read · with a concept loop
Separate Structure From Tuning Before You Trust an AI Design
Data, adaptation and alignment · 5 min read · with a concept loop
Multiscale Credit Assignment: The Credit Integrity Test
Agentic AI in production · 5 min read · with a concept loop
Multimodal AI Needs an Evidence Ledger, Not Just Bigger Models
Data, adaptation and alignment · 5 min read · with a concept loop
The Language Routing Contract: Localization Beyond Fluency
Agentic AI in production · 5 min read · with a concept loop
The Delegation Contract: When Multiple AI Agents Earn Their Cost
Model architecture and training · 4 min read · with a concept loop
The Three-Ledger Method for Model Architecture Decisions
Model architecture and training · 5 min read · with a concept loop
The Four-Clock Model for Mixture-of-Experts Costs
Context, retrieval and memory · 5 min read · with a concept loop
Memory Engineering: How AI Systems Should Govern What They Remember
Agentic AI in production · 5 min read · with a concept loop
Loop Engineering: How to Control When an AI Agent Stops
Operations and economics · 5 min read · with a concept loop
LLMOps and AgentOps: How to Release AI Agents Safely
Model architecture and training · 5 min read · with a concept loop
Latent Reasoning and Recurrent Depth: What Hidden Loops Are Worth
Agentic AI in production · 5 min read · with a concept loop
The Transfer-Compatibility-Confidentiality Test for Agent Memory
Data, adaptation and alignment · 5 min read · with a concept loop
Document-Mediated Policy Adaptation: The Bounded-Edit Ledger
Data, adaptation and alignment · 5 min read · with a concept loop
The Co-Design Ledger: Proving a Joint Harness and Weight Gain
Evaluation and assurance · 5 min read · with a concept loop
The Four-Claim Model: What AI Interpretability Actually Proves
Data, adaptation and alignment · 5 min read · with a concept loop
Why a Small AI Circuit Is Not Proof You Understand It
Strategy, product and value · 5 min read · with a concept loop
Why AI Agents Need Separate Goals, Metrics and Permissions
Context, retrieval and memory · 5 min read · with a concept loop
The Substrate Gate: Where AI Systems Should Store Information
Operations and economics · 5 min read · with a concept loop
Why Tokens Per Second Is the Wrong AI Speed Metric
Evaluation and assurance · 5 min read · with a concept loop
The Bounded Verdict Ledger for Auditing AI Research Claims
Responsible AI and governance · 5 min read · with a concept loop
The Four-Condition Test for Real Human Oversight of AI Agents
Agentic AI in production · 5 min read · with a concept loop
The Runtime Ladder: How Much AI Agent Infrastructure to Build
Evaluation and assurance · 5 min read · with a concept loop
The Manifest Ledger: Proving AI Development Actually Compounds
Responsible AI and governance · 5 min read · with a concept loop
Governance Engineering: How to Prove an AI System Is Compliant
Model architecture and training · 5 min read · with a concept loop
The Four-Gate Audit For Geometry-Aware Training Optimizers
Strategy, product and value · 5 min read · with a concept loop
The Novelty Gate: Testing AI Research Claims Before You Fund Them
Data, adaptation and alignment · 5 min read · with a concept loop
The Adaptation Gate: When Fine-Tuning Earns Its Keep
Data, adaptation and alignment · 5 min read · with a concept loop
Fast Weights and Test-Time Training: A Leader's Guide to Model Memory
Strategy, product and value · 5 min read · with a concept loop
Auditing an AI Program When Every Piece Passes Its Own Test
Data, adaptation and alignment · 5 min read · with a concept loop
Training Environments Need a Reward Contract, Not a Guess
Evaluation and assurance · 5 min read · with a concept loop
The Evidence Contract Behind a Defensible AI Acceptance Decision
Data, adaptation and alignment · 5 min read · with a concept loop
Stop Trusting Agreement Scores From Your AI's Own Grader
Evaluation and assurance · 5 min read · with a concept loop
The Evaluation Chain of Custody for Trustworthy AI Claims
Context, retrieval and memory · 5 min read · with a concept loop
Epistemic Trust Engineering: Why AI Needs a Time-Aware Evidence Core
Agentic AI in production · 5 min read · with a concept loop
The Four-Layer Envelope for Embodied AI and Robot Simulation
Small, specialised models · 5 min read · with a concept loop
What AI Model Compression Actually Changes Before Release
Model architecture and training · 5 min read · with a concept loop
The Hidden-Sequence Audit For Diffusion Language Models
Data, adaptation and alignment · 5 min read · with a concept loop
The Dataset Contract: Governing What a Model Learns
Strategy, product and value · 5 min read · with a concept loop
The Boundary-First Decision Atlas for Enterprise AI Systems
Operations and economics · 5 min read · with a concept loop
The Four-Denominator Ledger for Pricing AI Work Correctly
Data, adaptation and alignment · 5 min read · with a concept loop
The Reacquisition Clock: Testing Whether a Model Can Relearn
Data, adaptation and alignment · 5 min read · with a concept loop
The Update Ledger: Deciding How a Model Should Change
Context, retrieval and memory · 5 min read · with a concept loop
Context Engineering: How AI Systems Decide What They See
Evaluation and assurance · 5 min read · with a concept loop
Computational Provenance: How to Trust What an AI System Claims
Strategy, product and value · 4 min read · with a concept loop
The Core-and-Edge Model for Trustworthy Compound AI Systems
Model architecture and training · 5 min read · with a concept loop
The Intervention Test Behind Every AI Causal Claim
Model architecture and training · 5 min read · with a concept loop
Byte-Level and Hierarchical Models: A Leader's Guide to Text Units
Agentic AI in production · 5 min read · with a concept loop
Browser Agents That Prove an Action Actually Happened
Evaluation and assurance · 5 min read · with a concept loop
Why a Formally Checked Proof May Not Prove What You Meant
Responsible AI and governance · 5 min read · with a concept loop
Automated Alignment Research: Five Gates Before You Trust It
Model architecture and training · 5 min read · with a concept loop
The Reach-Record-Prove Method for Attention Architecture Decisions
Data, adaptation and alignment · 5 min read · with a concept loop
Reinforcement Learning Alignment: The Three-Signal Ledger
Agentic AI in production · 5 min read · with a concept loop
The Four-Gate Proof Loop for Governing AI Coding Agents
Agentic AI in production · 4 min read · with a concept loop
What It Takes to Trust an AI Research Agent's Findings
Data, adaptation and alignment · 5 min read · with a concept loop
Stop Paying for Information That Cannot Change a Decision
Agentic AI in production · 3 min read · with a concept loop
The Future of Agentic AI: Beyond Traditional Automation
Subscribe via RSS.