Create the Future
of Intelligence

AER Labs is an open-source artificial superintelligence institute advancing open science, efficient architectures, and the foundations of intelligence.

We build foundational understanding of intelligence to make it more open, more rigorous, and more widely accessible.

About

Intelligence is the substrate of civilization. It has to be open.

AI capabilities have advanced rapidly, but our understanding has not kept pace. Frontier systems are powerful, but often opaque. Knowledge of how they are trained and improved is concentrated inside a small number of labs. And while these systems may transform science, engineering, and society, the foundations of intelligence itself remain poorly understood.

AER Labs exists to close that gap. We aim to make artificial superintelligence more open, more understandable, more efficient, and more scientifically grounded.

Open Science

Scientific progress is a collective effort. We believe the path to artificial superintelligence should be inspectable, reproducible, and shared.

We plan to publish research, code, models, benchmarks, tools, and technical notes openly whenever possible. Open systems allow researchers, builders, institutions, and the public to verify claims, find failures, improve methods, and build on one another's work.

Closed systems create dependency. Open systems create science.

ASI-Native Research

AER Labs is designed for the ASI era. Traditional research institutes were built for a world where human cognition was scarce. In the ASI era, the basic unit of research changes. At AER, we see the new primitive as:

human researcher + frontier model + open codebase

This loop allows research to compound. Every theorem, model, dataset, benchmark, and tool should become part of an open substrate that accelerates future discovery. A result should not only be published. It should be runnable, inspectable, reusable, and capable of improving the next generation of work.

Acceleration, Efficiency, Recursion

AER is built around three principles.

Acceleration — AI can expand the supply of scientific cognition and help attack problems that were previously limited by human cognitive throughput.

Efficiency — Superintelligence should not depend only on larger models, larger clusters, and larger energy budgets. It should come from better principles, better architectures, and better representations.

Recursion — The institute must use its own outputs as inputs. Open tools and models should return as accelerants for the next cycle of research.

Together, these principles define AER's institutional form: an open, compounding research system for artificial superintelligence.

Foundations

We believe intelligence is not the same as knowledge. Knowledge is the product of intelligence. Intelligence is the mechanism that acquires, compresses, organizes, verifies, corrects, and improves knowledge over time.

Today's models contain enormous knowledge, but true intelligence requires self-correction, closed-loop learning, structured memory, and the ability to improve through interaction with the world.

AER Labs studies intelligence as a scientific subject: through information, compression, representation, prediction, feedback, and world models.

White-Box Intelligence

The future of AI should not be built entirely on black boxes. We are interested in systems whose architectures, objectives, representations, and learning dynamics can be understood from first principles.

Interpretability is not only a safety feature. It is a scientific requirement. A system that cannot be inspected cannot be fully trusted.

Human Discovery

AER Labs is focused on artificial superintelligence as a tool for scientific and civilizational progress. We are especially interested in systems that help researchers reason, prove, simulate, build, verify, and discover.

The goal is not simply more automation. The goal is deeper human-AI collaboration that expands what humanity can understand and create.

Building in the Open

We are scientists, engineers, and builders working toward open artificial superintelligence. Our goal is to build models, tools, infrastructure, and theory that make intelligence more understandable and more widely accessible.

Industry can make AI bigger. AER Labs exists to help make it open, rigorous, efficient, and right.

Open Source

55+ merged PRs across 30+ open-source projects the field runs on.

Fellows
Events

All events and seminars →

Join

The most important institute of the next decade is being built now.

Apply for Fellowship Sponsor