Make frontier intelligence small enough to own.

The largest AI models are extraordinary, and largely out of reach. They live in datacenters, behind APIs, metered by the token, governed by someone else’s terms. The most important question in applied AI is no longer can a model do this? but can you run the model that does?

Distillation is the answer. It is the craft of teaching a small model to think like a large one, transferring not just answers but reasoning, judgment, and nuance from a giant teacher into a compact student you can run on a laptop, a phone, or a single GPU. It is how intelligence escapes the datacenter.

What this is

The AI Distillery is the knowledge centre and research hub for model distillation and the applied edge of running AI yourself. We document how distillation actually works, from first principles to the methods being invented now, and we publish original research from models we distill and run ourselves. Accuracy is the standard: real numbers, primary sources, reproducible method.

Why now

Distillation was once a niche trick for compressing classifiers. Then reasoning models arrived, and with them the discovery that you can distill entire chains of thought, not just outputs. A small student trained on a frontier model’s reasoning can now rival systems many times its size on the tasks that matter to it. The field is moving fast, and it deserves a serious place of record. This is that place.

Documented, then built

The research here is not theoretical. It feeds real work: distilled GTM-operator models, quantization and hosting experiments, and the Stillhouse, where we distill specialized models for companies and host them behind an API. What we learn in production is documented here; what we document here sharpens the work.

Futuristic, but deeply human

There is something almost mythic about distillation. We take a vast, inscrutable intelligence and condense it, like a cloner drawing life from a template, into something a single person can hold and run and understand. The point is profoundly human: to put capability back in the hands of individuals, to keep your data yours, to make the future of AI something you participate in rather than merely subscribe to.