A First-Principles Augmented Intelligence System
We are happy to announce Logos, a first-principles augmented intelligence system.
Logos is in early development, but the results so far have been insightful across domains. Over the next few weeks we will be sharing a series of results of increasing importance in physics as the first domain. Today's is a lovely one that has been sitting quietly in special relativity for 121 years.
AGI
The dominant picture of where AI is going is one of autonomous capability at civilisational scale. Dario Amodei has called it a "country of geniuses in a datacenter." Sam Altman has written that "2026 will likely see the arrival of systems that can figure out novel insights." Elon Musk has said AI will be "smarter than all of humanity collectively" by 2030.
It is a compelling and somewhat terrifying picture. But if you look at how innovation has actually advanced, this isn't quite the picture.
Every major scientific breakthrough of the last century, from general relativity to DNA to the transistor to the Higgs, came from a small number of people seeing something nobody else had seen. Not from brute force across a large library of existing knowledge. From a single new angle that reorganised everything else. What current AI has demonstrated is remarkable talent: vast recall, pattern recognition, standard techniques at superhuman speed. What it hasn't yet shown is what Schopenhauer called the other half of the distinction.
"Talent hits a target no one else can hit. Genius hits a target no one else can see."
Demis Hassabis has called this jagged intelligence: systems that can win gold at the International Maths Olympiad but fail elementary maths if the question is phrased differently. Extraordinary in some places, blank in others, and the missing pieces are the ones that produce genuine breakthroughs. Humans fill the jagged edge, and that is how the hardest problems are going to be solved.
Logos
Intelligent Internet was founded to build tools for the Intelligence Age, set out in The Last Economy. The transition ahead will ask humans to reason about situations no previous generation has faced, at a pace no previous generation has navigated. First-principles thinking is how humans do that. It is also one of the few capabilities that generative AI doesn't yet have. The models have read everything humans have written about situations that have already occurred. They have no more experience of what is coming than we do.
To help us navigate what is coming, we went back to the foundations of intelligence to build a true first-principles reasoner. This required several breakthrough innovations to marry the best of human genius with machine talent. Logos is not a fully autonomous system, so the best benchmarks are breakthroughs. There is significant work to do, but the path is promising.
Physics
Genuine foundational progress in physics has been elusive for decades. Most proposals to move it forward require new postulates, new mathematics, or new physics.
Our results are different. They share two properties worth flagging:
- They require no new axioms or postulates.
- The mathematics is undergraduate level.
Today's shows that Einstein needed one postulate, not two, for special relativity. The mathematics has been available since 1888. The companion blog post sets the argument out in full.
More will follow in the coming days. The first major result lands on Wednesday, as a two-page paper.
What we are doing here is closer to Mythos finding the 27-year-old vulnerability in OpenBSD, or Lee Sedol's divine move in game four against AlphaGo. That move won him the only human game of the series, and was itself inspired by what he had seen AlphaGo play in Move 37. The tools were all already there. What was found was the opportunity.
We believe we don't need autonomous AGI swarms to solve the fundamental problems of the universe. Humans plus augmented intelligence designed to bring out the best in us is enough.
Innovation and intuition remain uniquely human. With the right tools, we believe that will be enough to unravel the mysteries ahead.
You can find our results on our website, follow on X, and keep up to date with Logos here.