Your blog description
Artemis II lifts off from Launch Complex 39B, April 1, 2026. Image credit: NASA/Michael DeMocker. I’ve been extensively testing Fable 5 (Mythos class) after the recent launch. My tests pushed its capabilities from spatial reasoning, critical reasoning, long running research and feedback, multi-step tasks, cinematography / visual design, and other more minor tasks. Examples: help research, design,…
If you’re like me and you ingest media compulsively it feels like we’re standing at a tipping point with release of Mythos . This is either the greatest marketing strategy ever (“so powerful we can’t let you use it”) or we’ve made another leap on our acceleration to AGI (see my other article ). My guess is with the release of GTP 5.5 with comparable Terminal-Bench 2.0 results at 82.7%, it’s all…
The unexamined life is not worth living. - Socrates This might make me the worst / most-SF person at every party, but my favorite question a few drinks in is to ask “what do you care about?”. The right time to ask is after two beers - first beer is to loosen everyone up, second is to get the honest answers. The first look I get is confused. People usually tilt their heads as if to say “Matt -…
Algorithmic progress: In the coming decade, AI labs will invest tens of billions in algorithmic R&D, and all the smartest people in the world will be working on this; from tiny efficiencies to new paradigms, we'll be picking lots of the low-hanging fruit. We probably won't reach any sort of hard limit (though “unhobblings” are likely finite), but at the very least the pace of improvements should…
Consumers are N=1 Makers Don’t be a consumer, it’s doesn’t lead to a life worth living. We hear all the time about the strength of the American Consumer, or the consumer economy, but what actually is a consumer? Not in the obvious “a consumer consumes” sense - a consumer is someone who buys things of course, but how we label our economy as driven by consumers. When you start looking, you can find…
Setting up an Nvidia GPU enabled system is the ante for deep learning research at home. Without a dedicated GPU, users are forced to either use Google Colab, which has great performance for a free service but is a notebook environment, or a cloud ML service that costs money (~$0.9 GPU/hour on AWS). After playing with those options for awhile, I decided to build my own system. In doing so, I…
The Demand: smart people who want lucrative jobs correlated with "smartness" I still don't get crypto, but I kind of get one of the needs for crypto. If you run in my " elite " 1 post-grad circle you'll notice that a lot of people talk about finance. Every quantitatively oriented software engineer I meet will introduce themselves and be like "I'm John. In my free time I like hiking, hanging out…
This post is going to cover the state of deep learning in 2021 with particular emphasis on AI infrastructure and deep learning tooling. If you're coming from a theory-based university classroom, get ready for an exhausting amount of detail. The classic pattern of "just feed some data through the network and backpropogate" is still the truth, but it takes ~10x more effort beyond the network to get…
When we talk about type safety in the context of statically typed languages, we mean that the language builds in a typing checking mechanism into the compiler. No type-check, no compilation. But in the context of research / algorithms oriented software that has no uptime requirements, the bias coming out of academia is to get the math into code, these days Python, and put it in production. Python…
Motivation This was for the Wikipedia competition hosted by Kaggle competition to predict future page view counts on Wikipedia. Like so many of the solutions now the winning approach used RNNs to learn a set of hyperparameters on the entire dataset that are then fed in to a predict locally on each time series. This begs comparison to multivel hierarchical models from a Bayesian perspective, but…
"Fighting with a large army under your command is nowise different from fighting with a small one: it is merely a question of instituting signs and signals." -Sun Tzu Overview and Common Assumptions The tactical lesson of the 2016 presidential election is that only a fool would blindly trust the polls. Listening to both sides' campaign managers the night before the election it was clear that…