Advanced Nonlinear Technologies Ltd.
Chaos theory, nonlinear dynamics, and the mathematics of self-organisation hold a secret: complexity doesn't require complicated causes. A single equation. A simple rule. Infinite, breathtaking structure.
We've spent two decades understanding this deeply enough to build with it.
Most fields move fast by ignoring what doesn't fit. We moved slowly by sitting with the things that didn't. These are the insights that took the longest — and proved the most durable.
Every system that looks impossibly rich — a coastline, a market crash, a neural firing pattern — has a generating rule that fits on one line. The hard part isn't the complexity. It's finding the rule underneath it. We've been building tools to find rules for two decades.
x₍ₙ₊₁₎ = r · xₙ · (1 − xₙ) → infinite structure
Working with Leon Chua on memristors changed everything we thought about how systems learn. Memory isn't a cabinet. It's the shape of a path through state space — a trajectory that bends the future. Biological intelligence exploits this. Current AI architectures mostly don't.
dH/dt = −λH + σ(Wx + b) → learning without forgetting
The human brain runs on 20 watts. A hummingbird navigates 3,000 miles on a body weighing less than a nickel. Biological intelligence doesn't brute-force its problems — it compresses, recurses, and self-organises. Sequential nonlinear processing is how nature does the impossible cheaply.
A decade spent studying how people learn revealed something startling: curiosity isn't random. It has topology. Questions cluster, bifurcate, and cascade — just like dynamical systems. The person asking "why" is a strange attractor tracing its own phase space. Understanding this changes how you build any system meant to help humans think.
Too much order: the system calcifies. Too much chaos: it falls apart. The most adaptive, creative, and resilient systems — immune systems, economies, great teams — all operate at the phase boundary between the two. Not by accident. By design. That's the only place where computation is truly powerful.
Ant colonies solve the shortest-path problem with no map, no leader, no plan — many of them blind. Each ant follows one rule: move toward stronger signal, leave signal behind. Intelligence emerges from the trail. Not from any individual.
We think learning works the same way. Conversation is the pheromone. The strongest ideas attract the most traffic. Dead-end trails fade. A robotic ant — indistinguishable from its peers — doesn't teach. It maintains the gradient. And the colony gets smarter without knowing why.
ANT is not just an acronym. It is a description of the architecture.
stigmergy / local rules → global intelligence / the medium is the memory
We are at an inflection point. AI has scaled to the edge of what brute force can do. The next leap will not come from bigger — it will come from understanding.
For thirty years, the field made enormous progress by ignoring the mathematics of natural intelligence and just throwing compute at the problem. That worked. Astonishingly well.
But the wall is visible now. And the researchers who will break through it are the ones who went and studied the other thing — the thing nature figured out in deep time. Nonlinearity. Emergence. The mathematics of systems that think cheaply and adapt fast.
We've been in that room for twenty years. We know what's on the other side of the wall.
Emergence over engineering
The most powerful systems don't get designed top-down. They get seeded with the right rules and then allowed to grow. Our work is about finding those rules — in biology, in mathematics, in the behaviour of agents.
Depth over velocity
We'd rather spend five years understanding something properly than ship something shallow in five weeks. The insights that took the longest have proven the most valuable — and the most defensible.
Structure over scale
Adding more parameters to a bad architecture doesn't fix the architecture. The hard, important work is finding the structural insight — the one that makes the problem smaller, not just louder.
The logistic map is one line of mathematics. It models how populations grow, how chemical reactions oscillate, how diseases spread, and how the heart decides to fibrillate.
Drag the slider — or hover anywhere across the diagram — and watch a single number take you from perfect stability, through rhythmic oscillation, through period-doubling cascades, into beautiful, deterministic, infinite chaos.
This is the shape of the problem we've spent our lives inside.
BIFURCATION DIAGRAM — where does x end up? (all time collapsed)
LANGUAGE SERIES — meaning at each value of r
drag slider · watch meaning drift into chaos
every word is real · the sentence is losing its mind
Research doesn't end. It deepens. These are the threads we're pulling — each one a convergence of what we've learned so far and where the mathematics points next.
Not by pruning a big model down, but by designing from the start for nonlinear sequential processing — the way biology does it. A system that reasons deeply on a fraction of the energy because the structure carries the load, not the parameters.
Active Research
Memristor dynamics give us a physical model for memory that doesn't catastrophically overwrite itself. The question is whether that principle can be translated into software architectures — and whether it unlocks the kind of lifelong learning that current models fundamentally cannot do.
Building Now
Wolfram's work on cellular automata shows that complex, coordinated global behaviour can emerge from agents following purely local rules. We're applying this to AI orchestration — systems that adapt their own tool use and task sequencing without being told how.
Theoretical + Applied
A decade of studying human learning suggests curiosity isn't noise — it's a structured signal with topology you can map. If that's true, it might be the most important input signal for any educational AI. And possibly for any intelligent system that needs to explore rather than exploit.
Long-Horizon Research
Chaotic systems are unpredictable but not unstructured. They have invariants — Lyapunov exponents, basin boundaries, recurrence patterns. Financial markets, immune responses, and creative cognition all use deterministic chaos productively. We want to know when AI systems should too.
Open Question
Everyday Series is our laboratory for this question. An App Store for AI agents built on nonlinear processing principles — developers compose once, businesses deploy at will. The platform is live. The deeper question is what new economic and social structures emerge when automation becomes this composable.
Live · Evolving
Domains where nonlinear thinking changes everything
"I've never been able to separate the beauty of a strange attractor from its usefulness. To me they're the same thing — and that intuition has turned out to be right every single time."
In 2005 I read Wolfram's A New Kind of Science and lost about three weeks to it. Not because it answered questions, but because it made me realise I'd been asking the wrong ones. The right question isn't how complex can a system get? It's how simple can the rule be while still generating everything?
I went to Budapest for my PhD specifically to work at the intersection of neural networks and chaos theory, when the received wisdom was that these were separate fields. My thesis was about systems that behave like strange attractors — circuits that don't settle, they orbit. I spent two years trying to disprove Leon Chua's memristor equations. He was right. We co-authored. The work made the IEEE cover.
After the PhD: IoT platforms, analog computing hardware, two US patents, a Royal Academy of Engineering fellowship, a developer community I built from nothing to 20,000 people. But the thread running through all of it was always the same question: why does intelligence compress so effortlessly in nature, and so poorly in machines?
The 20 years weren't a detour. They were the answer accumulating. Every collaboration, every failed hypothesis, every system that surprised me — it all pointed to the same structural insight about how nonlinear processing makes intelligence cheap. We're now at the point where that insight is buildable. That's exactly what we're doing.
Formation
Collaborations
Recognition
Built
If you're a researcher, an investor, a builder, or just someone who has spent time at the edge of a hard problem and felt the pull of the structure underneath it — we'd genuinely like to hear from you.
We're not looking for everyone. We're looking for the people who find this kind of thinking irresistible.