A free algebra will write anything at all.
What survives its admissions is the object.
Start with freedom. A free algebra generates without restraint — every word its symbols can spell, every product its grammar permits, every sum and every graded twist of sign, an unbounded country of expressions multiplying outward with no fact to answer to. Most of what it writes is about nothing. It is the widest thing in mathematics and the emptiest, a language standing ready before there is a world for it to be true of. Everything that follows is subtraction. The object we care about is what remains standing after the freedom has been made to answer, one demand at a time, for what it means.
A suspicion clings to any algebra built for a single purpose: that it has been rigged, its generators chosen to reach a wanted answer, its relations bent to yield the theorems its maker hoped for. The suspicion mistakes the whole history of the subject. William Kingdon Clifford, in 1878, assembled an algebra to hold quadratic forms and turned out to have built the algebra of spinors. Hermann Grassmann, decades before him, working alone and almost unread, put together the exterior algebra to carry the geometry of extended magnitudes and handed physics the differential form. The universal enveloping algebra was cut for the representations of a Lie algebra; the BRST complex — Becchi, Rouet, Stora, and independently Tyutin — for the ghost fields that keep a gauge theory honest; the antifields and the master equation of Batalin and Vilkovisky, in 1981, for the gauge theories that resist every simpler treatment; the Virasoro algebra for the symmetries of a string; the operator algebras of Rudolf Haag and Daniel Kastler, in 1964, for the observables of a quantum field, one algebra assigned to every region of spacetime. Each began as a bespoke thing, a formalism cut to the shape of one stubborn problem. Universality, where it came at all, came later, and to only a few, and usually as a surprise. Bespoke is the ordinary birth of nearly every algebra that ever mattered.
Our own algebra stands in that line. It is assembled from the common stock: tensor words and the graded signs that order them, the Koszul rule that carries a sign through every transposition, the BRST and Batalin–Vilkovisky machinery, the classical master equation, ghosts and antifields, the long grammar of what is closed and what is exact and the cohomology that measures the distance between them, quotients, shuffle and product corrections, transport, and the residuals left over when a field equation is asked to hold. None of that is ours; a mathematical physicist would read down the list and recognize every term, then look at what we did with those terms and reach for a longer name. The alignment laid over the common stock is what we made: a generator vocabulary cut for our subject — a set of active, paired BV-derivative generators — a bespoke product correction, a quotient that stays stable under both it and the BV differential, and an interpretation in which the surviving words carry physical weight, a laminar field, a source, a transport, a partial differential equation continued off a measured surface. What that longer name comes to is a custom BV/BRST construction with a domain-specific generator set, a quotient hierarchy, and an intent — to certify. The intent is much of its identity. It is an algebra whose reason for existing is to decide the moment at which structure read off the boundary of a living body has earned the name of a physical observable.
A bespoke algebra earns the right to be taken seriously the way any research algebra does, by carrying its weight: generators one can name, relations one can write, operations that close, theorems that are not obvious, obstruction surfaces one can locate, a quotient whose semantics hold consistent as you climb it, a reason fastened to every restriction, a map back into known mathematics and a map out into physical meaning. Ours earned it in the least romantic way imaginable, by being refused. The question underneath the whole construction is whether a product of observables stays physical — whether multiplying two admitted things yields a third that closes under the differential the algebra runs on. Multiplication is generous; it throws off branches in every direction, canonical field-and-antifield pairs, bare residuals, same-operator residuals, prefixed terms, tails of every length, and a lone term sitting at the tensor unit. The algebra reached, again and again, for a single clean law that would sweep them all up at once — one fold-completeness holding across every word — and the proof assistant, holding the whole thing in its kernel, declined to certify it. At the tensor unit there is no letter for the signed insertion to strike; the recursion that governs every interior word has nothing there to act on, and the term that lives at that edge has to be accounted for on its own. The machine would not let the boundary be smoothed into the interior. The true law came out narrower, and being narrower it came out truer: fold-completeness across the nonunit words, and the unit's contribution written down explicitly for what it is, a boundary term, the two together making a boundary-corrected whole. What looked from outside like a formalism a person had rigged to reach a wanted answer was, from inside, a person told again and again what he was forbidden to say. The shape of the algebra came from its refusals.
What the proof then showed, branch by branch, is how a product actually closes, and the how is far richer than the slogan that it does. The canonical pairs close by balancing inside their own package, a bare residual holding its value until the matching same-operator residual arrives to sum with it to nothing — each field paired with its antifield through the same table of active pairs. The nonunit tails close by a plainer grace: the carriers that would raise the forbidden coefficient run too long to reach it, filtered out by the length of their own support, so the dangerous term is never written and never has to be cancelled. Even the odd diagonal cases that looked most dangerous close that same quiet way, by support. And the lone term at the unit, once written down for what it was, was carried forward into the transport and shown to fall to zero once the surrounding structure left it no admissible contribution to make, so the edge blocks nothing downstream. The boundary was disposed of by a real condition, represented and paired and killed in the open, the quotient left unspent and the interior left unmerged. At each of these turns the proof was offered a shortcut — quotient the obstruction, declare the inconvenient thing equal to zero, move on — and at each turn it declined, because a quotient is built to carry a real physical equivalence, and spending one to hide an algebraic inconvenience would be a lie told in a second language. Nothing here closes by wish. Each branch closes because it is canonically paired, or excluded by the length of its support, or accounted for at the boundary, or routed through a residual the algebra already admits.
So the law that governs a product is the entire account of how it closes: canonical cancellation, support exclusion, boundary correction, same-operator transport, every branch made to answer in its own way. A composite observable earns the name physical only when each branch the multiplication threw off has been paired, or excluded, or accounted for at the edge, or admitted through a residual the algebra already holds. The same shape governs the instrument. A packet's features earn the name of a claim only when each has been witnessed at its source, calibrated, gated against motion, checked for phantoms, closed under the differential, carried faithfully through transport, and accounted for at the boundary. And beneath all of it lies the one result the proof kept uncovering, the one that belongs to this whole enterprise more than any theorem's name: the boundary is where the algebra decides what is real.
Step back from the single object and the same discipline stands up as a staircase. The free algebra, generating everything, sits on the floor. Above it a physical algebra keeps only what respects the form of physical law, and names that keeping admissibility. Above that an observable algebra keeps only what closes on itself, whose residuals actually fall to nothing, and names that keeping closure. Above that a measurement algebra keeps only what some real capture, taken from a real surface with its provenance sealed in, actually grounds, and names that keeping witness. At the top a claim algebra keeps only what the evidence beneath it has grown old enough to say aloud, and names that keeping maturity. Freedom on the ground floor, warranted claim at the ceiling, and between them a flight of admissions, each step a gate the level below must pass to reach the level above.
Formal methods usually climbs two of these stairs and stops. It narrows the free terms to the well-typed, and the well-typed to the provable, and there, at the edge of what is true in a model, it sets down its tools and calls the work finished. This staircase keeps climbing. It carries the same motion — carve the level above out of the level below with one admission — straight across the seam where mathematics touches the world, up through the measured and into the claimable. The instrument that lets it cross is the one already doing the carving: a claim's right to be made becomes one more predicate, checkable like the rest, and the question of whether we may say a thing turns from a matter of nerve into a matter of proof.
The words that name the gates are older than this use and carry more than one meaning at once, and the staircase holds together because at each step the meanings fuse. Closure is an algebra closed under its operations, and a set of statements closed under deduction, and a field equation that closes with its residual gone and its law satisfied. Witness is the object a constructive proof must exhibit to make an existence real, and the record a measurement leaves to make an observable real, one word for the single act that turns a possibility into a fact. Maturity is a grade of evidence and a readiness to be trusted. None of these is a pun. Each is one idea that happens to reach from logic through physics into the clinic without changing its nature, and the ladder is a single structure because the words that build it were already whole.
One stair differs from the others, and it rewards standing on. The first three levels are algebras in the full sense, subalgebras closed under their operations, so that anything built from admitted pieces is itself admitted. Witness and maturity behave otherwise. Add two witnessed observables and the sum is a stranger; the parts were measured and the whole never was, and the algebra's easy closure gives out. The top of the staircase is a filtration of evidence laid over an algebra, a grading loose enough that combination no longer stays inside it. The exact stair where combination stops being free is the stair where the object leaves mathematics and enters measurement. The break is the boundary itself, and the honest thing is to draw it where it falls.
All of this already bears weight in the instrument. Closure is the test a sealed reading is put to when its reconstructed field is measured against the templates of physical law and its residual reported honestly, whatever it comes to. Witness is the packet, the capture with its conditions and provenance sealed in, the record that turns a computed quantity into a measured one. Maturity is the discipline that holds a coefficient at zero until the predicate that would license it has passed, so that the instrument never says of a body more than the evidence taken from that body's surface will bear. The algebra and the certification are one apparatus, running from a torch and a lens laid against living skin all the way up to a claim a person can stand behind.
Which leaves the question that matters most, and honesty keeps it open. This algebra was cut to the shape of one instrument reading one kind of boundary. Whether it is bespoke to that alone, or whether it is the right calculus for boundary-generated observables wherever they arise — a cell's surface, an organ's face, the edge of a watershed, the photosphere of a star — no one yet knows, and we would be lying to say otherwise. There is one hint, and only a hint. The construction is scale-agnostic in a checked and literal sense: the same law and the same proofs survive when the underlying dimensions are resized, the rescaling maps commuting cleanly with the dynamics, so the object never leans on the size of the world it was born in. An algebra indifferent to the scale of its home shows the first sign of one indifferent to the home. Clifford's algebra came to matter because its shape kept reappearing, in spin, in relativity, in geometry he never lived to see; Virasoro's because a string was only the first place it surfaced. An algebra becomes important on the day its structure is found somewhere its author never went. Whether ours is waiting in other boundaries, in other bodies, at other scales, is the most interesting thing we do not know, and we hold it there, unclaimed.
This is the anatomy of the object, the account of how it is built and how it is made to answer. Whether the thing it computes actually tracks the health of a living body is a different question and an empirical one, still open, still being asked against real captures, and nothing here settles it. A bespoke algebra can be built beautifully and aimed at nothing. We have reason to think this one is aimed well. The proof of that lies in bodies, and we will say it only when the bodies have said it first.
A companion note, What is genuinely new, situates this object against the fields nearest to it; The making of an observable takes up whether the discipline it enforces is bespoke to one instrument or the shared grammar of observation. On what it does and does not prove, and the empirical work underway, see the science.