Iliad incubates technical research bets: theoretical projects that generate concrete, testable predictions, and methodologically mature, theoretically-informed experiments whose results feed back into theory.
Timaeus explores how capabilities and values emerge during neural network training by using interpretability tools built from principles from algebraic geometry and statistical physics combined with interpretability research.
Timaeus was launched before Iliad came into existence as an organization. Iliad’s Executive Director, Alexander Gietelink Oldenziel, was one of its cofounders — and Timaeus’ successes have provided a model for Iliad’s incubated orgs.
Simplex develops and applies a rigorous theory of latent internal structure in neural networks — how they internally organize representations, and how that structure relates to computation and behavior. They aim to build a theory of intelligence applicable to both artificial and biological forms.
Simplex was also founded before Iliad began as an organization, and Iliad’s Executive Director was also involved in the founding of Simplex. Its success further validated the model of incubated research organizations.
Principia builds theoretical model organisms that are simple enough to be mathematically understood, yet capture the complex phenomena we see in neural networks. Using tools from physics, it aims to bring analytical insights from these models that can yield testable predictions about the inner workings and behaviors of intelligent systems, contributing to a deeper understanding of the principles that govern intelligence and alignment.
XOR Labs is developing a principled understanding of how agentic behavior emerges in AI systems, what capabilities and risks result from this, and how these risks can be mitigated. They study how agency emerges in both individual AI systems and systems of interacting agents.
Stormglass studies how design choices shape model behaviors, uncovering the causal mechanisms of training itself. By developing rigorous theory, practical safety tools, and contributing to educational resources, it empowers researchers, labs, and policymakers to anticipate risks, detect unsafe behaviors earlier, and guide AI toward outcomes that benefit society. Stormglass’s work bridges disciplines, balances open science with responsible stewardship, and upholds scientific independence to help ensure future AI systems can be built on a robust safety foundation.
AIXI Labs is the leading mathematical model of artificial superintelligence, representing the maximum theoretical limit of AI capabilities. It models AI risk factors and safety mitigations in terms of AIXI variants, and develops the means to translate them to real AI agents. This enables rigorous testing of both the risk factors and the safety mitigations.
A peer-reviewed journal for foundational research in AI alignment. Launching in fall 2026.
Iliad incubates technical research bets: theoretical projects that generate concrete, testable predictions, and methodologically mature, theoretically-informed experiments whose results feed back into theory.
Iliad provides fiscal sponsorship and general operations support to incubated orgs. This includes administrative, budgeting, legal, and fundraising assistance, among others. When possible, we also provide seed funding.
Current incubated projects include AIXI Labs, Principia, Stormglass, XOR Labs, and the Alignment Journal.