Starting points for discussion. We write about AI, defence logistics, supply chain architecture, and what we’re learning along the way.
The Dark Data Centre
Buy the year’s NVIDIA output, keep it powered off, and let no competitor use it. Rent the racks back when the five-year depreciation clock forces it.
Read Post →What Does AI Sovereignty Mean to a Country?
Part one of a series. Before arguing about who should own a country’s AI, it helps to define what is actually being claimed.
Read Post →What Does “Agentic AI” Even Mean?
The term of art in 2026. No agreed definition.
Read Post →The Cost of 10 + 10
A solar calculator does it for nothing. A hosted LLM lights up several thousand tokens to reach the same answer.
Read Post →Robot AI Surgeon. What Day Is It?
wp.confdetail.subtitle
Read Post →The Monthly AI Budget
Tokens, not subscriptions. Per employee. With a refund for coming in under.
Read Post →The Model That Holds Still
Reproducibility is predictability — and for a commercial buyer that is worth more than a higher benchmark.
Read Post →The Child With the Credit Card
The spending decision was never the model’s to make.
Read Post →Don’t Ask the Model What More Compute Costs
The one source you can’t trust on the cost of more computational power is the model built by the company that bills you for it.
Read Post →You’re Paying to Fight the Phone
The real labour and subscription cost of locking down a consumer device for SME corporate use — and what you get for it
Read Post →The Case for a Boring Work Phone
Why fleet devices for company use should do less — and why that is a security posture, not a cost cut
Read Post →A Gate With No Fence: Why Robots.txt Was Never Going to Work
The honest case for a DNS-based AI crawl protocol — and a clear-eyed admission of what it cannot stop
Read Post →The Vampire Squid Effect: AI Is Extracting the Web It Needs to Exist
Why the economic contract that built the open internet is broken — and who is going to pay to fix it
Read Post →Voting LLM Systems: Lessons from NASA for AI Reliability
How Multi-Model Consensus Prevents Hallucinations in Mission-Critical Business Applications
Read Post →Multi-Motor AI Strategy: Matching AI Response to Physical Inertia
Matching AI Response to Physical Inertia in Industrial IoT
Read Post →The Entropy Paradox: Why Eliminating Hallucinations Will Kill AI Creativity
Why the Drive Toward Perfect Accuracy Produces the Most Expensive Dictionary Ever Built
Read Post →Python and the AI Toolchain: How Ecosystem Gravity Replaced Language Design
Why a Slow, Interpreted, 30-Year-Old Scripting Language Became the Infrastructure of Modern AI
Read Post →The Reliability Gap: Six Sigma Standards and the AI Uptime Problem
Why the Metrics That Governed Telecoms for Decades Have Not Yet Been Applied to AI Services
Read Post →The DevOps Assumption: Why AI Services Break the Model That Made Cloud Software Reliable
How AI Services Inherited DevOps Vocabulary While Discarding Its Discipline
Read Post →Trained Past Useful: Overtraining and the Usability Cost of the AI Arms Race
How the race to benchmark supremacy is producing models that score higher and work worse — and what that means for the people building on top of them.
Read Post →