Auckland, New Zealand
You signed up at kinarey.com. If someone forwarded this to you, welcome — you can subscribe at the bottom.
There's a principle in nutrition: the diet that reverses chronic disease is the same one that minimises environmental damage. Not two fixes. Just one.
I walked over 2,000 kilometres last year. Trying to match it this year — but I'm behind, and 6.6 km a day, every day till December, is daunting. The deeper I get into the health side of things, the more I notice this pattern everywhere: problems that look separate keep turning out to be the same problem.
It happened again last month. Enterprise leaders worrying about AI cost and control. Environmentalists worrying about the grid — 448 trillion watt hours and counting. Two conversations, zero overlap. But the architecture that makes an organisation resilient is the same one that cuts its energy consumption by orders of magnitude. Not two fixes. One. That's the essay →
Hit reply if anything here lands — I read every one.
If you find any of this useful, forward it to someone who'd appreciate it.
The Efficient Frontier
A family in Maryland gets a utility bill spike from a data centre three states away. A CTO in London watches token costs fall and total spend double. A policymaker in Geneva reads the numbers: more electricity than all but ten countries. Three people. Three cities. The same crisis — and an architecture that resolves all of them at once.
Read →The Gravity Well
Token prices are falling. Total spend is rising. That's not a paradox — it's the oldest trick in platform economics. The unit of measurement shifts quietly beneath you while the dashboard still shows green. This essay tracks how that sleight-of-hand works, and what it means for anyone budgeting AI spend into next year.
Read →Who Gets to Ask the Slow Questions
$2.5 trillion in AI spending this year. The world's first trillionaire. And the governance frameworks meant to guide all of this? Principles documents that nobody is obliged to follow. Research on 500-year-old monasteries reveals the mechanism that actually works — and it's not what the tech industry is building.
Read →The cost of confidence
A new research paper coins the term “cognitive debt” — applying the economics of financial bubbles to AI adoption (arXiv). The argument: stability breeds overconfidence, overconfidence breeds fragility. The more capable the tool, the less visible the dependency — until it breaks. Meanwhile, more than half of enterprises report watching AI agents fail confidently, producing wrong answers with no signal that anything went wrong (VentureBeat). The evaluation gap is widening faster than the capability gap.
Architecture as intervention
Alibaba’s SkillWeaver framework cut agent token consumption by ninety-nine percent — not through better models but through structural routing (VentureBeat). Instead of loading an entire toolbox for every request, it routes tasks to composable skill modules. The most impactful AI efficiency gain right now isn’t a chip or an algorithm. It’s plumbing.
Governance without teeth
New Zealand’s AI framework for government gets called “relentlessly optimistic” — strong on aspiration, thin on enforcement (The Conversation). Meanwhile, China’s agentic AI policy takes the opposite approach: humans must stay in the loop, full stop (The Register). Two countries, two philosophies, same underlying question: who enforces the principles once the document is signed?
And then there’s this
AI solved an 80-year-old combinatorics problem that mathematicians had been stuck on since the 1940s (Nature). A reminder that when these systems are pointed at the right problem — bounded, well-defined, genuinely hard — they remain extraordinary.
You're receiving this because you subscribed at kinarey.com.
If someone forwarded this, get your own copy.