Articles

Essays on systems that have to survive reality.

Architecture, governance, delivery, and the practical edge where AI stops being a demo and starts becoming an operating problem.

Article · October 2, 2026

The Model Is the Engine. The Race Is Won in the Pit.

A sourced map of the ten-layer AI infrastructure stack, what remains contested, and which architecture decisions European enterprises should keep reversible.

14 min read ai-infrastructure / enterprise-architecture / reversibility

Article · October 2, 2026

The Infrastructure Consequential AI Agents Need

AI agents that affect money, data, rights, services or physical environments need bounded authority, evidence and real user recourse.

7 min read ai-agents / accountability / digital-identity

Article · October 2, 2026

Consolidation as Historical Norm

How recurring bottlenecks in radio, telephony, cable and publishing illuminate concentration in AI infrastructure, and how open institutions can change the pattern.

5 min read market-concentration / communications-history / ai-infrastructure

Article · October 2, 2026

Day in the Life: 2031

A grounded thought experiment about one ordinary day with delegated AI agents, digital identity, runtime controls, health support and media provenance.

10 min read ai-agents / scenario / digital-identity

Article · October 2, 2026

The Great AGI Illusion

AGI may transform machine capability. Access, authority, distribution and economic value will still depend on the infrastructure around it.

6 min read agi / ai-infrastructure / market-power

Guide · October 2, 2026

How Do You Test a Control That Can Never Fail?

A practical guide to testing safety and policy controls with mutation, path evidence and an explicit UNEVALUABLE result.

10 min read control-testing / runtime-governance / assurance

Article · October 2, 2026

How Open Source Saved the Internet

The internet stayed contestable through open protocols, shared software and viable commercial ecosystems. AI now faces a harder version of that test.

5 min read open-source / internet-history / ai-infrastructure

Article · October 2, 2026

The New Mainframe

Frontier AI has a concentrated, mainframe-shaped core. Its application edge remains more open, and that distinction will decide where power settles.

5 min read ai-infrastructure / market-concentration / computing-history

Article · October 2, 2026

The Three AI Stacks

The US, EU and China are building distinct AI systems around different rules, infrastructure and economic priorities. The seams matter most.

8 min read ai-policy / geopolitics / ai-infrastructure

Article · October 2, 2026

What Your AI Pilot Actually Proved

An AI model can pass a pilot while the surrounding operating system fails. Production exposes integration, workflow, lineage and runtime risks the demo never tested.

5 min read ai-operations / enterprise-ai / production-readiness

Article · October 2, 2026

Who Captures the Rent?

Who captures durable value in AI infrastructure and applications? History points to the scarce complement customers cannot easily reproduce, substitute or leave.

5 min read ai-economics / market-power / strategy

Article · October 1, 2026

The Model Can Be Probabilistic. The Authority Boundary Should Not Be.

Hard rules own actions an AI agent must never take. Typed semantic judgment handles bounded ambiguity, and qualified review sees the consequential residue.

13 min read runtime-governance / ai-agent-security / high-risk-ai

Article · June 13, 2026

Runtime Is the Moment of Consequence in AI

The Therac-25 record shows how fatigue, unsafe assumptions and a missing effective interlock converged at execution time. Some AI-agent action paths recreate parts of that risk.

11 min read runtime-governance / historical-analogues

Article · May 11, 2026

When AI Governance Reaches the Execution Boundary

Documentation matters. It becomes operational only when consequential actions pass through an enforceable boundary with reconstructable evidence.

5 min read runtime-governance

Article · April 11, 2026

AI Agent Approval Fatigue Is Real

If your coding agent asks for approval on every harmless action, people either turn it off or stop using the loop. The fix is not more prompts. It is a policy layer.

2 min read runtime-governance / ai-architecture

Article · April 11, 2026

The Missing Layer in Autonomous Agents Is Runtime Policy

Teams trying to productionize autonomous agents usually do not fail because the model is weak. They fail because there is no clean policy layer between the agent and the tools it can trigger.

2 min read runtime-governance / ai-architecture

Article · April 2, 2026

Why I'm Building an AI Operating System

Most AI tools are isolated workflows. I wanted to understand what happens when intent, execution, governance, and learning belong to the same system.

2 min read practice-notes / ai-architecture

Article · March 20, 2026

AI–Code Division of Labour

Reliable agentic systems get faster and more truthful when models stop doing deterministic work and start handling only ambiguity.

1 min read ai-architecture / runtime-governance

Article · March 8, 2026

Keeping Authoritative Data Outside Model Generation

A model cannot alter a field it never regenerates. That protects one authoritative data path, not the factual correctness of the whole system.

3 min read ai-architecture / runtime-governance