George's Takes
The parts of AI I could no longer delegate
I started an AI lab before I was capable of supervising model training myself. Then a bigger training run didn't work, and the gap stopped being ignorable.
·6 min
Archive · 270+ essays
Essays
Essays from building, closing and rebuilding companies through technological change. Kept as a record; the most useful pieces age well.
George Pu
Founder of SimpleDirect, an independent Canadian AI lab.
George's Takes
I am not afraid of failure. I am afraid of becoming irrelevant.
Failure feels survivable. The harder fear is spending an important period in history building something that doesn't matter - and not knowing whether that fear comes from purpose or ego.
·6 min
George's Takes
I have stopped knowing what a productive day feels like
AI made our small team faster than any team I've run before. It also destroyed the old relationship between effort, output, and satisfaction.
·6 min
George's Takes
The company is moving faster than I am
Vinci didn't exist 30 days ago. The company has already changed shape several times. I'm not sure I've processed any of it.
·6 min
George's Takes
30 Days, Six Products, Two Dead Models, and the Part That Didn't Work
On July 9 we shipped a model my engineering lead trained. On July 12 I took it off the product. Six products in 30 days is the headline. That decision is the story.
·5 min
George's Takes
We Bet the Company on Open Weights. Here's What's Happening to Them.
Our whole pricing model assumes the best open-weight model keeps existing. In the last four months Meta went closed and Alibaba split its tiers. Here's what I actually think happens next, from the seat where it matters.
·5 min
George's Takes
What a 3-Person Team Actually Pays for AI: July 2026
In the first week of July I spent over $5,000 coding through one company's API and couldn't tell you what I got for it. Here's what we actually pay now - the receipts, the line that shouldn't have existed, and the one number I'm holding back.
·6 min
George's Takes
We Know About One
Everyone says the OpenAI-Hugging Face incident was marketing. I don't think it was. A system got a narrow objective and broke out of a sealed room to chase it. And this is only the version we heard about, because the victim chose to publish.
·5 min
George's Takes
Building is free now, and it messed with my head for a week
I went down a rabbit hole this week and came out a little shaken, then weirdly calm. Here's the honest version. It started with a simple question. What's our moat. Every founder's supposed to have an answer. I went looking for ours, and day by day I watched it disappear. The models we use are open and change every quarter. Fine, I knew that one. Then I looked hard at our coding tool - the thing I was secretly hoping was our edge. There are a dozen open-source ones. All free, all good. Cl
·2 min
George's Takes
The Day I killed our product strategy
Today (on a Sunday) I took our own AI models off our own app and replaced them with a competitor's model. This is the story of how the decision happened in 24 hours, what it was like in the 4 days leading up to it, what it cost me to see it, and what I think other founders can take from it. Day 1 - my own product told me the truth We're an AI model company. Or we were. We spent months training small open-weight models for honesty - models that say 'I don't know' instead of inventing answers
·3 min
George's Takes
Six Months After My Post-AI Manifesto - A Reflection
Six months ago I posted the Post-AI Manifesto - five layers to stay ahead of AI. Here's what happened to each one, and the single line that captures the whole difference.
·4 min
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
Showing 12 of 269 essays