Essays

Long-form arguments on how technology decisions shape products, organisations and markets: AI architecture, edge platforms, and engineering governance.
Night landscape: a bright stream of releases runs from a campus to a city while a dimmer branch turns off toward a walled complex marked as the sovereign path

The Divergence Tax

A compliance boundary is approved as a certification and operated as a second release train. The cost nobody budgets is the divergence rate, and architecture sets it.

Machine-speed code generation flowing into a merge decision hub that carries generator, blast radius and named owner, with one path leading to production release and another to a postmortem

Verification Is Not Accountability

Verification tells you AI-generated code is sound. It cannot tell you who decided to accept the risk. The case for recording decision lineage in the merge path.

A near-future city at dusk with residential broadband devices, industrial facilities, a telecommunications tower, and a data-centre campus interconnected by glowing network lines — the physical infrastructure layer beneath the AI stack that vertical slices anchor into.

The Slice Underneath the AI Layer

Layer commoditization is now consensus. The harder question is what a defensible AI business looks like afterwards: a vertical slice fusing data and workflow to a device fleet, a jurisdiction, or a regulatory regime.

Accountability and AI-generated code

AI Didn't Break Accountability. It Exposed the Gap.

AI coding tools broke the accountability chain engineering organizations relied on for decades. Developers report 42% of the code they commit is AI-generated, and most haven’t decided who owns the outcome when it fails.

AI on Edge

AI on Edge: Autonomy

Edge AI is about autonomy and survivability, not just latency. When connectivity drops or round-trip time matters, the edge has to make decisions on its own. That changes how you architect AI systems and who controls them.