- AI with Alec Newsletter
- Posts
- AIWA "The One Thing" #24: Chamath Called It. Satya Nadella and Lin Qiao Confirmed It. The AI Coca-Cola is the Learning Loop.
AIWA "The One Thing" #24: Chamath Called It. Satya Nadella and Lin Qiao Confirmed It. The AI Coca-Cola is the Learning Loop.
Chamath Palihapitiya said this three years ago. He wasn’t wrong then. He isn’t wrong now.
“It’s like giving two great chefs three ingredients but you give the third chef one extra one. That person has the opportunity to do something very special.” - Chamath Palihapitiya, July 2023.
Last week, Satya Nadella and Lin Qiao arrived at the same place from different chairs.
There isn’t one frontier. There are many.
Frontier models are the refrigerators. Extraordinary technology. Real value. Not the endgame.
The labs will create some Coca-Colas. VC-backed AI-native startups will create more. But many of the most enduring ones will come from companies that bring the one ingredient nobody else has.
The Learning Loop Is the New IP of the Firm
The company that builds a compounding loop between human expertise and AI capability owns something no model release can replicate. You can offload a task. You can never offload your learning.
The moat isn’t the model. It’s the ontology, the domain expertise, the people’s taste and judgement. Your data and workflows. Enterprise Intuition doesn’t commoditize. Models do.
You know you’re cooking when you can swap out the generalist model without losing the expertise built into the system around it.
If not, you’re renting intelligence. Not owning it.
There Isn’t One Frontier. There Are Many.
Frontier models are infrastructure. But infrastructure is not the moat. The most interesting development isn’t how smart one model is becoming. It’s the way intelligence is being harnessed and becoming customizable.
A model post-trained on proprietary company knowledge, grounded in on-going decision traces and context graphs. That’s a frontier. A specialized system that solves one narrow problem better than anything else is a frontier. A router mapping the right token to the right task is a frontier.
Every company that encodes its institutional knowledge creates a frontier that belongs to it. The companies that win won’t have the biggest models. They’ll have the most specific ones.
The Gap Is Still the Game. Every New Frontier Widens It.
The deployment gap hasn’t closed. It’s growing. Every new model release creates more distance between AI capabilities and where most enterprises actually are.
The companies closing the gap right now are compounding advantages that become harder to replicate every month that their intelligent systems are being used in production.
The rising tide lifts ships that row. The architectural work: data infrastructure, knowledge layer, orchestration layer, learning loop. That’s what makes it work for you.
The enterprise opportunity is real. The window is not infinite.
The companies doing the architectural work today, the rewiring, are the ones who will look back and say they saw it coming. While sipping their Coca-Cola.
Humans + Machines. Never Humans vs. Machines.