About
Making AI systems you can verify
HeavyClaws is an AI safety and research company based in Manchester, UK. We build modular, verifiable, locally-deployable AI systems.
Our Purpose
The industry default is monolithic: one black-box model, one provider, one point of failure. We are proving that a different architecture is not just possible — it is already training. Modular AI that runs locally, keeps your data yours, and is verifiable by construction.
What We Do
Modular by Design
Independently-trainable experts — each verifiable, each auditable, each replaceable. The architecture enforces isolation. A failure in one component cannot cascade.
Safety Is Structural
Safety primitives built into the architecture — not filters applied after the fact. Local-first by design. Your data stays on your hardware.
Built for the Edge
O(1) memory complexity. Consumer GPU training. Systems designed to run on your hardware — the only architecture that respects the user.
From Manchester
No scaling games. No Valley VC narratives. Modular architectures trained on our own GPUs, under our own roof, with no external dependencies.
What We Value
Every decision we make is measured against our mission: building AI systems that people can verify, trust, and run on their own terms.
01
Put the mission first.
Nothing else matters if the architecture is wrong. Modular, verifiable, local AI is the only defensible path forward.
02
The simple thing that works.
O(1) memory is simpler than KV caches. Local is simpler than cloud. Closed-source is simpler than an open attack surface.
03
Safety is not negotiable.
Safety primitives are structural. We build systems that refuse — not systems that apologise.
04
Built in Britain.
Manchester-based. Registered in England and Wales. No hype cycles. Just engineers building things that work.
Governance
HeavyClaws Systems Ltd is a private company registered in England and Wales (company no. 16998646). Based in Manchester, UK. All research, training, and deployment happens on our own infrastructure.