Day 0 — a technology and research project, built from zero
XYMiracle is being built from zero. We are starting with AI compute infrastructure, with the question of what computing should become when it is designed around AI, and with how machines can build models of the physical world.
01Day 0
Most early projects describe the company they intend to become. Here is what is true instead.
02Where we’re starting
They are not a roadmap. Two are questions about computing; one is a question about how machines come to understand the world. Each one is a starting point with a reason to be there.
Use today’s technology to explore scalable AI computing infrastructure — and validate whether a sustainable compute business can be built on it.
We are exploring what it actually takes to provide computing to the organisations that need it: AI companies, startups, research teams, universities and other institutions. Power, networking, operations, utilisation and unit economics are all part of the question.
Today this is a direction to validate, not an operating cluster.
Learn from real AI workloads, and explore what future computing systems should become when they are designed around AI itself.
The goal is not another conventional GPU workstation. Today’s CPU, GPU and server architecture should not be assumed to be the final form of AI computing. We want to find where real workloads break, and whether new architectures, devices, accelerators, chips, memory systems, interconnects, cooling or energy approaches are what AI actually needs.
Today this is a question we are asking, not a design we have built.
Explore how machines can build deeper models of the physical and natural world.
World models are one promising direction, not a permanent commitment. The broader question is how machines come to understand nature — mathematics, physics, chemistry, biology, materials and engineering — through observation, hypothesis, prediction, simulation or experiment, comparison with reality, and correction.
Today this is a research direction, not a system we have built.
These are our starting points — not the boundaries of XYMiracle. What that means.
03The loop
Compute gives us capability, and we expect it to give us revenue. Real workloads then show us what is wrong with today’s systems. Fixing that is what makes harder research affordable — and harder research is what demands better systems again.
01
Practical computing capability — and, if we get it right, the first commercial revenue.
02
Real workloads show us where today’s systems actually fall short.
03
Research creates demands that existing hardware was never designed to meet.
04
Systems designed around those demands — in architecture, memory, interconnect, cooling or energy.
05
Better computing makes more ambitious questions affordable to ask.
This is a direction of travel, not a roadmap with dates — the loop closes over years rather than quarters. Each step depends on the one before it working commercially and technically, and we have not proven any of them yet.
04How we think
01
Ambition is a starting condition, not a result. An idea that has not survived evidence, experiment, engineering and real-world testing is still a hypothesis, and we will treat it as one.
02
Exploration requires independence, and independence requires resources. We intend to build commercial capability that pays for long-term research — not to make money first and decide what to do with it later.
03
We hold our current directions loosely. Compute, world models and whatever comes after them are starting points, not conclusions — including the ones we are most excited about.
04
Day 0 is described as Day 0. If we cannot show it, we do not claim it — no matter how much better the other version would read.
05Join
We are not filling roles. We are looking for the first people who can change what XYMiracle is able to do — across AI infrastructure, GPU computing, distributed systems, networking, data centres, hardware, ML systems, research, business development, finance and supply chain.
If your background is not on that list but you can see a way to make this real faster, that is worth more to us than a matching CV.
XYMiracle is at the very beginning. The first people to join will help decide what it becomes.