Sovereign Compute
Why We Shipped Vinci on the Eve of Canada Day
Nobody takes Canada seriously on AI - and for years we earned it. The compute wall, why it finally fell, and why we shipped Vinci Piccolo on the eve of Canada Day.
·4 min
257+ Essays
Essays
Deep thinking on building businesses designed to own forever. Not how-to content. Decision logs, frameworks, and pattern recognition.
George Pu
Builds in AI. Still owns everything.
George's Takes
I Got Bored of Being a Commentator
I got good at having AI takes - and one day realized I'd become a commentator, not a builder. A complaint is just a wish with better production values. So I built the replacement.
·4 min
Own Your Tech
Why We Started With Character
Capability is getting cheaper every month. What's scarce is a model you actually want to talk to. Why we bet on character - and started with a small, open, honest 4B model you can run yourself.
·1 min
George's Takes
The Barrier to Building AI Isn't Talent. It's Capital.
Startups in San Francisco raise $18 million just to afford the compute. We did the same work bootstrapped. The real gate on who gets to build AI isn't ability - it's who can afford to be in the room.
·2 min
George's Takes
What It Actually Takes to Get a GPU in Canada
We had the money and we're a multi-year cloud customer. It still took two months, a flood of scammers, and a $14,000-a-week quote to rent a few chips. Here's how GPU access actually works.
·3 min
George's Takes
Anyone Can Copy the Methodology. Only One Lab Can Keep the Promise.
Vinci ships August 8 with the methodology published in full - so competitors can copy it. The thing they can't copy is whether the model actually does what its constitution says, in public, under tests anyone can run.
·6 min
George's Takes
What I Got Wrong About Canadian-Specialized AI
For six months I was sure the answer was a from-scratch Canadian AI model. Seven days ago that broke. I had confused the model with the system - and the open-book version is the bigger bet.
·6 min
George's Takes
Own or Be Owned: The Week a Letter Switched Off Fable 5
On June 12 a government letter reached past Anthropic and switched off the best coding model I'd ever used. The shutdown wasn't the lesson - what it revealed about renting intelligence was.
·4 min
Own Your Tech
We Are No Longer Building a Canadian Legal AI Model
Yesterday I published the post-mortem: we asked flash-1-mini ten questions any Canadian lawyer would consider basic, and it invented seven citations. That post was about what broke. This one is about what it changed. Because two weeks ago I was telling people we were building a Canadian legal AI model - and today we decided we're not. I want to walk through why, because the answer changed how I think about what "building AI" actually means. The original scope The original scope made sense
·6 min
Own Your Tech
We Asked Our 4B Legal AI 10 Questions. It Invented 7 Cases.
An honest post-mortem on our own model — and what anyone building or buying AI should take from it.
·8 min
Own Your Tech
Open Weights, Closed System
Google shipped local AI on the Mac this week. The weights are Apache 2.0. The experience is a walled garden. That's not a contradiction. That's the playbook.
·5 min
Own Your Tech
Why CBLRE Matters More Than the Model
Yesterday we released CBLRE — the Canadian Bilingual Legal and Regulatory Evaluation. The day before, we released flash-1-mini, a 4-billion-parameter bilingual Canadian legal AI model. Most of the launch coverage has focused on the model. That's the wrong artifact to focus on. The model is the proof. CBLRE is the moat. Here's why. The gap nobody had filled Before yesterday, no standard public benchmark existed for Canadian bilingual legal AI evaluation. That sentence is bigger than it so
·5 min
Showing 12 of 257 essays