Essays
Thinking out loud.
Not tutorials. Slower pieces on the things I keep turning over — AI and the shape it's giving the world, how we build and learn, the odd bit of economics or product thinking. I write these to ask better questions, not to sound certain.
Filter by topic below; multi-part reads are gathered into series. Shorter, build-focused posts live in Writing.
Not sure where to start?
Notes from the Winter
AI & Society4-part series · 54 min in total
- 01The Good WinterWe seem to be living through a Fourth Turning — a once-in-a-lifetime crisis season. Here is the case, built on data rather than dread, that it is the best possible time to build, learn, and help each other — and how to actually use it.
- 02Something in the AirThe uncanny feeling that ideas arrive to many of us at the same moment is real — and it is not mystical. A grounded tour of multiple discovery, the adjacent possible, the collective brain, synchronization, and what the algorithm is doing to all of it.
- 03Seeing Early, Thinking SidewaysA step-by-step, science-backed field manual for the two skills every founder secretly needs: spotting the ripe idea before the crowd, and reaching the one nobody else can see. With three thousand years of examples — from Vedic-era India to the modern feed — and a recipe you can actually run.
- 04More LightIn 1865 a young economist noticed something strange: making coal-burning more efficient made Britain burn far more coal, not less. That paradox is usually quoted as a warning. Read it forward into the age of cheap intelligence and it becomes, I think, the most hopeful thing you can say about what comes next.
Deep Roots
Systems & Tech3-part series · 30 min in total
- 01The First Bit Was a SyllableAround 2,200 years ago an Indian scholar named Piṅgala set out to count the rhythms of poetry — and on the way invented binary numbers, the Fibonacci sequence, Pascal's triangle, and a fast way to raise numbers to powers. Here is how counting verse built the logic your phone runs on, explained from scratch.
- 02The First Program Was a GrammarTwenty-five centuries ago, Pāṇini described the whole of Sanskrit as a formal system of some four thousand rules — a machine that generates a language. It is, by any honest reckoning, the ancestor of the programming language, the compiler, and the way we teach machines to handle words. Here is how a grammar became the first algorithm.
- 03The Oldest GameThe most ancient thing humans made that you can still sit down and do today isn't a tool or a text — it's a game. A cheerful romp from a 4,600-year-old board dug out of the ground to the day a machine played a move no human would dare, and what our endless love of play says about us.
The weight of a thought
Every clever answer an AI gives has a physical cost — electricity, water, heat, metal. Here is the real footprint of modern AI, why training is only half the story, and how progress and a livable planet might actually share a future.
The Programmable Network
Systems & Tech3-part series · 12 min in total
- 01Your phone network is quietly getting a brainFor a century it was a dumb, reliable pipe. In one decade it learned to take instructions. In the next, it starts to think. Here's the whole arc — no telco degree required.
- 02The network grew superpowers (and handed you the remote)Knowing where a device really is. Sensing that a SIM was just swapped. Promising a fast connection on demand. These used to be telco magic. Now they're a function call — and there's a 'universal remote' making it work everywhere.
- 03When AI holds the keysGive an AI agent API keys and every superpower becomes a door it can open — fast, tireless, and at machine scale. What I took from apidays, why most breaches are gloriously unexotic, and how not to get robbed.
The economy runs on stories (and now the machines write them)
Robert Shiller's uncomfortable idea: markets move on contagious stories, not just fundamentals. Add knowledge representation, the randomness of what goes viral, and the algorithms now amplifying it all — and you get the real operating system of the modern economy.
Learn to fall early
We are born fearless problem-solvers — thousands of falls a day, and we call it learning to walk. Then we spend years training it out. A case for taking real risks early, while they are still cheap.
Who is AI actually for?
The most interesting question about a technology is rarely what it can do. It is who it is for.
We keep asking if AI is smart. I keep asking if we are being wise.
Intelligence is a property of the machine. Wisdom is a property of the people building it. We have spent far more on the first.
What we lose if we stop struggling
AI is brilliant at handing us the answer. I keep wondering what happens to us when the struggle to find it goes away.