Summary In today's episode, I explore how people worldwide can reduce their dependence on US-based technology companies by tapping into sovereign AI labs, local inference providers, and open-source models. Here's what this means for you. You gain a practical roadmap to access capable AI that runs within your own jurisdiction, keeps your data private, and…
“Ask me up to 20 clarifying questions until you have enough information to succeed at the task.” I have been advising folks for years to tell AI to ask it questions, but I’ve noticed something in the most recent generation of models – any model that’s tuned for agentic AI specifically really needs a kick…
Pangram, the AI that powers the new Substack AI validation system, is proud of its 0.02% false positive rate. Let’s frame that appropriately. First, call it what it is: a false accusation rate. In any situation that matters (not Substacks, but things like students being kicked out of school for academic dishonesty, professors losing tenure,…
China strikes again. Two big model releases this week – Minimax H3 (Hailuo), king-size video generator, which can run on a machine that can support 500 GB of VRAM. You won’t be running this on a MacBook any time soon, but big inference providers like DeepInfra will be able to serve this up and it’s…
Two enduring maxims: Fast. Cheap. Good. Choose any two. You get what you pay for. Earlier today, my friend alerted me about a discussion on X, visible in the photo on this post. TLDR: people said they didn’t want to pay for books because they couldn’t afford them, and thought all authors should be required…
