Attention is everything
The science
of directed
intelligence
We build AI systems that know where to look – and why. Research at the frontier of attention, reasoning, and machine understanding.
Explore our workThe problem with intelligence has never been scale. It's been selection – knowing which signals matter, which questions to hold, which connections to trust. We are building machines that attend with precision, reason without drift, and learn with the economy of understanding.
Research Philosophy
Mechanisms Before Scale
We study how intelligence actually works – in circuits, not parameters – before deciding how to grow it. Understanding precedes capability.
Efficiency as Elegance
The most powerful systems are precise, not large. We design architectures that do more with less – fit for edge, fit for the world.
Research That Ships
Every paper we write, we also build. Ideas earn their place by surviving contact with real hardware, real latency, real use.
Intelligence is not about knowing more. It is about attending to what matters – and having the discipline to ignore everything else.
Attenor Research Charter
Focus Areas
Attention Architecture
Novel hybrid attention mechanisms that go beyond transformer defaults – combining recurrent compression, selective gating, and associative memory into coherent, trainable systems.
Mechanistic Interpretability
Opening the circuit beneath the weights. We map what models actually learn – the features, the routing, the emergent structures – so capability and understanding grow together.
Edge-Native Intelligence
Frontier-quality reasoning on a laptop. Architecture-first efficiency – not quantization as an afterthought – designed from the ground up for constrained hardware.
Reasoning & Multimodal Systems
Models that hold chains of thought without drift. Systems that move across language, vision, and signal with the same precision – not bolted together, but unified by design.
We are a small team building something that will not be small.
Stay tuned.