Work with us
One budget. Built into the machine, at every layer.
See the stack, or come build a layer with us.
ethan@admissibletech.com
A quantum state is a set of distinctions — this and not that — held against a world that keeps trying to erase them. Holding each one costs something real, the supply is finite, and distinctions crowded together compete for it. Read the machine that way and its hardest problems stop looking separate. Decoherence, crosstalk, the error-correction wall, the trust gap — each is the same budget overspent in a different place.
Every piece of quantum information is a distinction the hardware has to keep straight: a phase that has to stay this and not that, long enough to compute with. Maintaining that distinction has a real, ongoing cost — preventing premature collapse takes continuous work, and information is physical in the most literal sense: erasing it costs energy, measured, not assumed. Any finite region of the world can only hold so many distinctions at once before they blur into one another. That ceiling is the budget. A processor is an exercise in spending it well.
Picture a sheet of paper that holds only so much writing before the lines start to smear. Add more, press them closer, and the page stops being legible — not because any single mark is wrong, but because the page ran out of room. A qubit fails the same way, and for the same reason.
Pull on any of the field's hard problems and the same thread comes out: a budget being spent faster than it can be held. So the defense is not five unrelated fixes. It is one budget, protected five times — and we hold a patent at each layer.
Each layer has its own page in the stack.
None of the ingredients are new. Information has been known to be physical since 1961. The bound on how much distinguishable structure fits in a region is decades old. Error correction has had thresholds for thirty years. So why hadn't the machine been read as one budget before? Three reasons, none of them mysterious.
First, the failure modes live in different departments. Decoherence belongs to physics. Crosstalk belongs to microwave engineering. The threshold belongs to coding theory. Trust belongs to metrology. Four fields, four vocabularies, four sets of fixes — and no shared currency in which anyone could notice they were the same overspend.
Second, the reflex is to add, not to budget. When a quantum machine struggles, the instinct is more — more qubits, more redundancy, better materials. Treating capacity as finite and allocatable, as a thing whose spending you design, cuts against the habit of solving scarcity by buying more of it.
Third, the lens came from a different building. Seeing one budget took starting from another question entirely — what does a finite world actually permit? — and only then looking down at a processor. The people best placed to fix quantum hardware were not standing where the whole ledger is visible at once.
We did not discover that information is physical, and we make no claim to have. We took a limit the field already trusts and treated it as a working budget rather than a curiosity at the edge — then built the defense layer by layer. The pieces were on the table the whole time. The move was to read them together.
What follows is the longer version — the mechanism rather than the metaphor — for readers who want it.
A distinction is a finite, enforceable separator between possibilities: a boundary the hardware draws and then has to keep drawn. Holding one draws a cost against a finite capacity. Moving between two distinguishable states costs a positive floor that the universe sets and will not discount — Landauer's kBT ln 2 to erase a bit, the Margolus–Levitin limit on how fast a state can change at a given energy. These are not engineering inefficiencies to be polished away. They are the cheapest the world allows. The budget is the running total of those unavoidable costs against the ceiling, and a quantum processor lives or dies by how close it runs to the line.
Maxwell's equations can be written in terms of the fields, or in terms of an underlying four-potential. The potential carries more components than the fields strictly need, so the standard formulation fixes a gauge — it sets some of those components to convenient values and works with the rest. It is an efficient and entirely standard choice. What Aharonov and Bohm showed in 1959 is that the potential is physical all the same: a charged particle picks up a measurable phase traveling through a region where the magnetic field is exactly zero, where only the potential is there to act on it.
None of this requires a new electrodynamics, or a claim that the potential is more 'real' than the fields. The point is the plain one Aharonov–Bohm already settles: a description can carry real, measurable structure in the parts a convenient choice sets aside. The capacity budget is structure of exactly that kind — present in the physics all along, just not the quantity the standard accounting was built to track.
In the language of the equations: the gradient of the four-potential has sixteen components, the electric and magnetic fields are built from six of them, and a gauge choice sets the rest aside — a standard, useful simplification, and the kind of place where real structure can go unnoticed.
The magnetically-dark qubit (AT-002) works in exactly this regime, on purpose. A coaxial solenoid confines the bias field so that B ≈ 0 outside while the potential does not vanish — the Aharonov–Bohm configuration, used as a design tool. Cancel the field a neighbor can feel; keep the control you need. The deep idea and the shipping hardware are the same idea, one layer apart.
The principle is grounded in the company's technical papers and the mechanisms are worked out analytically and confirmed in simulation and field solvers. The performance figures are theoretical and validated in those tools — not yet proven in silicon. The vacuum layer is an open frontier, and we say so.