We earn commissions when you shop through links on this site — at no extra cost to you. Learn more

246+ Essays

Essays

Deep thinking on building businesses designed to own forever. Not how-to content. Decision logs, frameworks, and pattern recognition.

George Pu
George Pu
$10M+ built. Still own everything.

Not sure where to start?

My Best Thinking
George's takes
Frameworks
How I think
My Portfolio
Real numbers
About George
The full story
Why CBLRE Matters More Than the Model
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
Closed Orchestrators Will Commoditize. Open Ones Will Compound.
George's Takes

Closed Orchestrators Will Commoditize. Open Ones Will Compound.

Two AI infrastructure announcements landed in the same week. Ours was a model. We released flash-1-mini — a 4-billion-parameter bilingual Canadian legal AI — under Apache 2.0, alongside an open benchmark (CBLRE), an open training corpus, and the methodology behind both. The model runs on a MacBook. It runs on a Raspberry Pi. The weights are downloadable. You own the file. Perplexity's was an orchestrator. At Intel's Computex keynote in Taipei, CEO Aravind Srinivas demonstrated what the com

·5 min
What We're Building: An Open-Weight Canadian Model Series

What We're Building: An Open-Weight Canadian Model Series

The model is the smallest part of the story. Here's what it is, what it isn't, and what comes next. Today we shipped flash-1-mini. It's a 4-billion-parameter open-weight model, fine-tuned for Canadian context, bilingual in English and French, that runs on a laptop with no cloud dependency. You can download it, run it offline, and own it. The weights are yours. I want to write about what it is, what it isn't, and what comes after — because the model itself is the smallest part of the story.

·6 min
While You Were Watching the Chatbots
Policy & Economy

While You Were Watching the Chatbots

Over the past two years, Canada quietly rebuilt the question of who controls its artificial intelligence. Not in one announcement. That's the point. There was never a single moment loud enough to make you look up. The decisions arrived in fragments — a Christmas Eve letter, a contribution agreement with a file number, a press release at a university most people don't follow, an MOU with the important parts blacked out. Each fragment was, on its own, small enough to ignore. Together they red

·5 min
Canada Is Renting Its Sovereignty
Own Your Tech

Canada Is Renting Its Sovereignty

In the last piece I wrote about the anchor: Canada designated Cohere as its national champion at the foundation model layer — $240 million, a 24-year term, an MOU that called the company "the only sovereign, cloud-agnostic large language model operating in Canada," and a fund Cohere itself proposed. That designation is settled for a generation. This piece is the opposite kind of writing. The last one was analysis of public records, and I tried to stay out of it. This one carries a point of

·8 min
How Cohere Wrote the Fund That's Now Funding Everyone Else
Policy & Economy

How Cohere Wrote the Fund That's Now Funding Everyone Else

On Christmas Eve 2024, the Government of Canada issued a Letter of Intent to Cohere for a Strategic Innovation Fund contribution of $240 million. The timing is the first thing worth noticing. Christmas Eve is when governments send letters they want done quickly and quietly. The Trudeau government was already wobbling. Chrystia Freeland had resigned on December 16, eight days earlier, in a public letter that called out the Prime Minister's "costly political gimmicks." Less than two weeks lat

·8 min

Free Tools

14 calculators and assessments. Runway, valuation, AI-proof score, and more.

Browse Tools
Sovereignty Is Not About Ownership
Own Your Tech

Sovereignty Is Not About Ownership

A vice-president at a Canadian university called me last month. Her president had told her the institution needed to be on sovereign AI within the year. The provincial government had announced a funding program. Other universities were already applying. She wanted to know what she should be evaluating. I asked her what sovereign meant in this context. She paused for a long time. Then she said, "I think it means the data stays in Canada." I asked what about it staying in Canada specifically

·6 min
I Almost Told a Lawyer to Build His Own AI
Own Your Tech

I Almost Told a Lawyer to Build His Own AI

A lawyer friend — call him Mark — called me this week. He and another friend had spent the weekend trying to run an 8 billion parameter language model on a 16 gigabyte laptop. Mark thought he was going to show his friend something impressive. The output was gibberish. Incoherent strings of text that no junior associate would have signed off on. He'd come to me because he wanted to know what hardware to buy next. On the call, I didn't know. What I did know was NVIDIA's software stack. I'd

·6 min
Fine-tuning your own AI doesn't cost $35,000. It cost us about $50.
Own Your Tech

Fine-tuning your own AI doesn't cost $35,000. It cost us about $50.

Two A100 graphics cards. Spinning quietly in a Google datacenter. Five hours of training. About $50 in compute. That's what it cost us to fine-tune our own 4-billion-parameter AI model this week. The base model went from 30% accuracy on the tasks we care about to 98%. Read any article on fine-tuning costs and you'll see numbers between $5,000 and $35,000. One blog called it a 'CFO conversation.' Another listed 'hidden expenses' that could double your initial estimate. A third quoted team

·5 min
Your ChatGPT and Claude Conversations Are Court Evidence
George's Takes

Your ChatGPT and Claude Conversations Are Court Evidence

Greg Brockman's journal became Exhibit 161 this week. The next chapter writes itself. Someone's ChatGPT history becomes Exhibit 162. That sentence sounds like speculation. It isn't. The infrastructure is already in place. The court orders are already in place. The only thing missing is a famous enough defendant for the headline to break the way Brockman's did. The court order most people haven't read In May 2025, Magistr

·6 min
The Journal That Exposed the President
George's Takes

The Journal That Exposed the President

Greg Brockman keeps a diary. Not a metaphor. A literal, decades-long, type-into-his-laptop journal that he started in college when he was deciding what to study. He kept it through Stripe. He kept it through founding OpenAI with Musk and Altman. He kept it through the brawl for control of the company in 2017 - the fight that's now being relitigated in a San Francisco courtroom. This week, his journal became Exhibit 161. Hun

·4 min
One Rack Is a Cloud
Own Your Tech

One Rack Is a Cloud

What colocation is, and why most AI founders have never heard of it

·10 min

Showing 12 of 246 essays