Data science leader, local AI builder, open-source maintainer.
To build anything with local models you need to use them as coding agents. In this post, I talk about my open-source side-project - minimal-ai - that simplifies all of that into a few steps.
I recently resuscitated my old blog that I started back in 2008, which led me to write this very different post for my Substack — hopefully a good mix of fun, nostalgia and sarcasm for my readers.
I've been writing about local AI all year. The biggest consumer tech company in the world presenting it on their main stage is a very positive development for open source & local AI.
In my quest to continue testing out local AI models that don't need data centers, I gave them a live A/B test to analyze and see if they can replace me in my daily job
Sharing my experience from testing local AI models in a practical way, using visual prompts to build something you can see and assess quickly.
An easy setup guide to help you run a private AI on your own laptop in under 15 mins
I've enjoyed the ride with Claude & Codex. But rising costs, quality regressions, and the rise of capable local models might bring private and trustworthy AI to everyday users soon.
Earlier in the year, I used AI to build me a prompt analyzer app in 20 mins. It took me weeks to wade through the slop and apply the right methods to make it work correctly.
A workflow for building real things (not demos) with AI coding agents
The victory of solving a hard problem was sweeter when things used to take time. Now, getting a dopamine hit from watching an agent do all the work, is not the same at all.
Frontier models are extraordinary. On-device alternatives are not there yet for the average person. But the privacy and cost math is shifting fast.
I analyzed a year of my Claude conversations. The patterns were more revealing than the content itself.
Most data scientists inherit someone else's pipeline. I built one end-to-end — product design, instrumentation, statistics — to show how it's done at the highest levels.