Writing
How AI is reshaping decisions, organizations, and the systems we rely on — the second-order effects nobody talks about. Published when ready.
These essays reflect my personal perspectives only and do not represent the views of my employer.
0
essays
0k
words
0
topics
Essay series
Series · 6 parts
The Efficiency Stack
TurboQuant, ChatJimmy, Sora, converging models, and the model-as-menu: why the next phase of AI is decided by delivery economics, not IQ.
Start with part one →
Series · 5 parts
Second-Order Effects
What becomes newly valuable once intelligence gets cheap — jobs, judgment, atoms, the deciding mind, and a frontier gone open.
Start with part one →
Start here
Second-Order Effects: The AI Jobs Story We're Missing
Everyone asks what AI will replace. The better question is what becomes newly valuable once intelligence gets cheap.
The Efficiency Stack · Part 6
OpenAI didn't ship a model this month. It shipped a menu — three tiers of the same release on one cost curve. That is the efficiency thesis becoming the org chart of the industry.
Second-Order Effects · Part 5
Frontier-class weights are now free to download. The first-order read is that the labs' moat is gone. The second-order read is that it moved somewhere harder to copy.
Second-Order Effects · Part 4
AI collapses the cost of producing options. It does nothing to collapse the cost of choosing between them. That gap is where organizations will quietly break.
Second-Order Effects · Part 3
Intelligence is getting cheap faster than anything physical can keep up. The next decade belongs to whoever can make atoms move at the speed of bits.
Second-Order Effects · Part 2
AI is eating the work that used to train people. The efficiency gain shows up this quarter. The missing generation of judgment shows up in ten years.
The Efficiency Stack · Part 5
GPT-5.4, Claude 4.6, and Gemini 3.1 launched within weeks of each other. No clear winner on benchmarks. That is the most important result.
The Efficiency Stack · Part 4
OpenAI killed Sora because $15 million a day in inference costs dwarfed $2.1 million in lifetime revenue. It is the strongest case study yet for why delivery economics — not capability — decides what survives.
The Efficiency Stack · Part 3
Everyone asks how smart the model is becoming. The better question is how cheap intelligence is becoming to deliver.
The Efficiency Stack · Part 2
Most AI demos are framed as intelligence demos. ChatJimmy is more interesting as an economics demo.
The Efficiency Stack · Part 1
In many real systems, the scarce thing is not intelligence itself. It is working memory — the cost of keeping enough of the past alive for the model to be useful in the present.
Why AI is making presence the premium — and what that means for the future of strategy work.
Most leaders still talk about AI as though it were a better search box. OpenClaw suggests the bigger shift: AI as an ambient layer that can act wherever work already happens.
Second-Order Effects · Part 1
Everyone asks what AI will replace. The better question is what becomes newly valuable once intelligence gets cheap.
I went all-in on vibe coding for a real project. The speed was unreal. The debugging was a nightmare. Here's the honest scorecard.
People frame AI decisions as 'human vs. machine.' The real question is where to sit on the spectrum between them — and most organizations choose poorly.
Strategy decks are easy. The hard part is who retrains the model, who handles the 2am edge case, and who decides when to override the AI.
A killer demo gets funding. It gets press. It gets a standing ovation. What it doesn't get you is product-market fit.
I learned to read code as a strategy person who'd spent a decade outside engineering. Here's the mental model that made modern codebases legible.
Every day I watch autonomous systems make life-or-death decisions they can't fully explain. Here's the framework I use to think about trust when stakes are high.