Hyperplane is an AI research company.
We research cognitive architectures for AI memory and intelligence, drawing on principles from cognitive science to build systems that remember, reason, and learn.
We believe that memory is the core of cognition.
Our research program
Perception
Comprehension of new inputs. What it means, what matters, how it connects to what's already known.
Consolidation
Continuous reprocessing to extract patterns, merge redundancies, and let unimportant details decay.
Abstraction
Moving from specific instances to general knowledge. Concept formation over retrieval
Reorganization
New information restructures what's already known. Rethinking and learning based on new inputs.
Anticipation
Advance pattern projection and forward remembering. Knowing what to surface before it's asked for.
Reflection
Self-directed review of what the system knows, identifying gaps, contradictions, and decay.
Our research powers Exabase M-1, which achieved state-of-the-art on the LongMemEval benchmark using a smaller, cheaper model.
AI that serves humanity.
AI systems should augment human capability, not replace human agency. Memory architectures that are transparent, auditable, and human-directed are how we get there.
We believe the next generation of computing will be built on AI that understands us, protects us, and empowers us to do our best work.