Building the theoretical foundations of AI
We identify fundamental principles governing intelligent systems, both natural and artificial.
Each law has a precise mathematical statement with verifiable predictions.
Rigorous scientific methodology: testable hypotheses, experiments, and empirical evidence.
We map relationships between laws creating an interconnected knowledge network.
We don't seek verification, but the accumulation of avoided refutations. Each failed prediction brings us closer to the truth.
Language model optimization tools emerging from our theoretical research. Efficiency without compromise.
A living network of knowledge about intelligence
Continuous symmetries in cognitive systems imply conserved quantities analogous to those in theoretical physics.
We discovered fractal self-similarity in cognitive systems, and it passed every test we threw at it. The only law in our compendium with perfect validation confidence.
We analyzed 30+ embedding datasets. PC1 captures 88-99% of variance in every single one. Your high-dimensional vectors are hiding a low-dimensional truth—and it is costing you 40x in storage.
How we achieved 40x compression on vector embeddings while maintaining >95% search accuracy, reducing storage costs from $600/month to $15/month.
Interested in collaborating or learning more about our research?
From Santiago, Chile, we build the theoretical foundations of artificial intelligence. We collaborate with researchers worldwide.