When a program checks a claim and reports a pass, the pass means as much as the test behind it. This thesis argues that a verdict binds, or holds as a real result, only where its criterion is externally witnessed and re-derivable: someone outside the checker wrote the test, and anyone can run it again. Everywhere else, the record should say that no outside check backs the claim yet.
The studies come from Telos, the author's workbench for AI research, whose tools turn AI and physics talks into records a reader can check. A pass that nobody outside can repeat still looks like proof to the people who rely on it, and that is why the line matters. Four recent studies from this project, plus one added case, show where the line falls. Study 3 is the one case where a check passed and holds, and the closing section lists what is proven and what stays open.
Words used on this page
Verdict
The answer a checker gives about a claim. This project uses three verdicts. MATCH means the claim agrees with its criterion. DRIFT means the check ran and the claim disagrees with its criterion, or has moved away from it since an earlier pass. UNVERIFIABLE means no outside test backs the claim: none exists, none is in reach yet, or none has run, so the checker does not rule.
Judge
The separate checker that fills in a verdict.
Criterion
The standard a claim is checked against, such as an exact answer from standard math software.
Externally witnessed
Someone outside the checker wrote the standard, and others can see the check and repeat it.
Re-derivable
Anyone can repeat the check from the same inputs and get the same result, or watch it fail.
Binds
Holds as a real result.
Bound claim
A verdict that holds, because someone outside the checker wrote its criterion and anyone can re-derive the check.
Bid
A verdict with no such outside check behind it. Only the checker's confidence at the moment backs it. The page also calls it an unwitnessed bid.
Conferred
Granted from outside.
Aseity
A philosophers' word for standing on your own footing, with nothing outside holding you up. The conferred-existence thesis, which this page builds on, says nothing has it. This page applies that claim to verdicts.
Bounded rationality
Reasoning with limited tools and time. Here it means a checker can rule only on what its tools reach, as the seven-loop card in Study 2 and the finite viewer in Study 5 show.
Falsifier
A stated result that would prove a claim wrong if it turned up.
Hash and seal
A hash is a short fingerprint computed from a file. If the file changes, the fingerprint changes, so a reader can confirm a record is the one that was checked. A seal is a hash taken over a whole collection.
Forge
A pipeline in this project that turns talks and papers into records a reader can check. The Learning Forge and the Discovery Forge are two of them.
Claim card
One claim written down with its source and a slot for a verdict.
Simulate leg and render leg
The two parts of the C3 claim in Study 3. The simulate leg is the math, checked in software. The render leg is whether a real device does it.
Substrate
The project's own software and data.
The reconcile
The project's name for one checking routine with three steps: perceive a shape, check it against a criterion it did not write, and carry a proof that someone else can re-check. The reconcile thesis is the author's earlier written argument for that routine.
The argument in one paragraph
A made thing has no aseity: it has no footing of its own, and something outside it holds it in being. A verdict is a made thing of that kind. A MATCH binds where two things hold: someone outside the checker authored the criterion it was checked against, and anyone can re-derive that check. Outside that domain, the same verdict is an unwitnessed bid, backed only by present confidence. The Learning Forge, the Discovery Forge and a long research diet, a reading program of eight sealed talks, all stop at the same limit. Only one claim in either forge started as a bid and crossed into a bound claim: C3, a thermodynamic result. It bound on one leg only, the simulate leg, where a standard linear solve supplied an outside criterion. The render leg, the claim about a physical device, stays UNVERIFIABLE. The asymmetry marks how far each check could reach, and the records state that limit as it stands.
This page ties four recent studies to the reconcile thesis. The viable-visualization thesis, the author's study of whether a finite system can show mathematics faithfully, joins them as an added case, with a marker that Study 5 explains. Every non-trivial claim below carries a falsification condition or a citation to a source record in the underlying draft. Proof before trust applies to this thesis too.
The three claims
The thesis rests on three claims. Each one applies the claim before it one level further out.
Claim 1 · Conferred status has no aseity, and neither does a verdict.
The conferred-existence thesis, the author's paper on what it means for anything to exist, defends ontological nihilism, the view that nothing has being of its own, in its only survivable form: no aseity, the denial that anything is self-standing. Existence, status, standing, the moral ought (what a person should do) and legitimate authority all share one ontology, one account of what kind of thing they are. Each is conferred, relational and re-spoken each instant, which means it has to be granted again moment by moment. None has aseity, and none is any less real or binding for lacking it. The thesis turns this on itself. By its own anti-realism, the view that such statuses exist only because someone grants them, its central claims are bids that present power certifies until they are earned.
A verdict is a made thing of the same kind. When a verifier emits MATCH, that MATCH has no footing of its own. A criterion outside the verifier holds it in being. Remove the external check, or move to a domain with none, and the verdict is a bid. The link to the conferred-existence thesis is literal: the verifier shows the same structure, the separation of ground from conferral.
Claim 2 · A verdict binds precisely where its criterion is externally witnessed and re-derivable. Beyond that, it is a bid.
This is the reconcile thesis stated as a boundary. The reconcile is one operation: perceive a shape, check it against a criterion it did not author, and carry a re-checkable proof. The Discovery Forge record states a rule it calls the Discipline: a result must be SIMULATED (computed in a simulation) or derived, and then checked against a criterion it did not author. A result that is only RENDERED, meaning drawn or displayed, or only asserted, falls short.
For the verdict to bind, the criterion has to meet two conditions at once. Someone outside the checker wrote it, which keeps the checker from grading its own homework. It must also be re-derivable by an independent party, so the check can be repeated and can fail. MATCH is a bound claim only when both conditions hold; if either fails, the verdict is an unwitnessed bid and the correct output is UNVERIFIABLE.
This mirrors the conferred-existence thesis at its hardest point. There, the ought binds while a person inhabits the practical standpoint, the position of someone deciding how to act who cares whether a reason holds. The person can decline it by withdrawing from that standpoint. Read "criterion's domain" where that thesis writes "practical standpoint", and the modal structure, the logic of when a claim must hold, is identical. The bindingness and the boundary are both real, and the limit takes nothing from the ought's force inside the standpoint.
Claim 3 · The boundary is the honest shape of verification.
The Learning Forge's ungrounded modules and the Discovery Forge's UNVERIFIABLE convergence claims, which say the research talks line up with the project's design, mark the correct limits of present reach. The records state those limits plainly. The C3 result shows the contrast: a bound claim with its criterion present, a measured MATCH on the simulate leg, and the render leg still UNVERIFIABLE. The rest of this thesis gives the evidence for Claim 3, one study at a time.
Study 1 · The Learning Forge: empty verdict slots are the honest state
The Learning Forge turns frontier AI talks and papers into learning objects. A learning object goes further than a summary, because it carries its own evidence: a source with a hash, and a claim with a verdict slot that a separate judge can later fill with MATCH, DRIFT or UNVERIFIABLE. The collection of sources, the corpus, is sealed, and any reader can re-derive its inventory from the hashes in the cards. That part is witnessed and re-derivable.
All six claim cards carry a verdict slot left empty on purpose. The record gives the reason in one sentence: grounding a claim in a hashed source and adjudicating it are two separate steps. The forge gathers and cites, and a separate judge decides; until the judge runs, the slot stays empty.
This is Claim 1 in operation. A grounded citation carries no verdict. The card has a witnessed source, the hash, and no external criterion has checked the claim yet, so the card cannot bind. Filling the slot with MATCH before the judge has run would confer a verdict with no ground: a bid presented as a bound claim. The forge refuses, and the refusal is correct. The scoping note goes further. The forge covers ten modules, or topic areas. Five are backed by evidence, two are partial, and three have little or nothing on point and need sources gathered. The forge lists those three modules as ungrounded and leaves their coverage empty.
Proven
The corpus is sealed and re-derivable. Six cards are grounded in specific hashed sources.
Open (UNVERIFIABLE)
All six claim cards. No judge has adjudicated any of them. The record says so and leaves the slots empty.
What would prove this wrong
Re-run the corpus list and digest. If the inventory or the seals do not match, the grounding claim is refuted. If the module tally is anything other than five grounded, two partial and three ungrounded, the summary here is refuted. If a card's slot turns up filled with a MATCH that no judge produced against an external criterion, that card is a bid posing as a verdict and should revert to UNVERIFIABLE.
The Learning Forge is the cleanest case in the thesis, because no claim in it crossed from bid to bound.
Study 2 · The Discovery Forge: three falsifiers, and one verdict that became real
The Discovery Forge is an assembly line. It turns a research source into a discovery object that carries its own re-check. Its intake rule is the falsifier rule: the forge rejects any card that has no falsifier. That enforces Claim 2 at intake. A claim that cannot name a criterion it did not author, and a way that criterion could fail, could never bind, so the forge does not admit it.
The convergence-thesis mapping claimed that the research talks support the system's design, meaning the design of Telos. It went through the judge, and all six of its claims came back UNVERIFIABLE. The judge has no oracle, meaning no outside measure, that can tell whether a physics talk supports a design, so it refuses to assert a MATCH it cannot ground. Claims 1 and 2 meet here. The mapping is interpretive. No external, re-derivable criterion can adjudicate it, so the only verdict it can support is UNVERIFIABLE, and the verifier's refusal is itself the re-checkable artifact. The interpretation is faithful and labeled as interpretation. It carries no measurement, and no oracle exists to supply one.
The three falsifiers
The Discovery Forge stakes its discipline on three named falsifiers, each dated or computable. Each is a criterion the forge does not author.
Falsifier 1 · a dated neutrino-mass prediction (the neutrino card). If one right-handed neutrino is stable and is the dark matter, then the lightest neutrino is massless, which forces a predicted minimum for the sum of neutrino masses. The test is an external, dated measurement the forge does not control. Track the published upper bound on the sum of neutrino masses from galaxy-clustering surveys over a stated window of three to four years. If the measured upper bound drops below the stated minimum, the prediction is falsified. If the surveys detect a bump, a signal at that level, it is confirmed. This is the cleanest verifier target in the corpus, because it is dated and anyone outside can re-check it. Its verdict slot reads UNVERIFIABLE for now, by construction: the data for the window is not yet in hand.
Falsifier 2 · the thermodynamic stochastic process (the C3 result). C3 is the third card in the Discovery Forge and has its own page. Study 3 covers it in detail. It is the one falsifier that flipped to a real verdict.
Falsifier 3 · a seven-loop quadratic-gravity calculation (the seven-loop card). Quadratic gravity is a proposed extension of Einstein's gravity. In a certain limit, it embeds into a scalar field theory that is known to seven loops on that side. Quantum gravity in this limit is known only at one loop. Each loop order adds a finer correction to the calculation. Put plainly, physicists have already worked the related theory out to high precision, so the forge could check its answer against theirs once it has software able to do the algebra. The criterion is the independently computed seven-loop result, and each loop order is a checkable target the forge does not author. The verdict slot reads UNVERIFIABLE, because the system has no symbolic engine, the kind of software that does algebra, able to carry seven-loop perturbation theory today. The card is parked as a well-posed, externally checkable goal that is out of current reach, and it stays on the list.
A fourth card uses an experimental method as a template. It makes no physics prediction for the forge to measure. The card is marked interpretive, and its slot reads UNVERIFIABLE. The three falsifiers map onto the thesis like this:
Neutrino card
Binds in the future, when an external measurement arrives.
C3 result
Binds now, on one leg, because its external criterion, a linear solve, is available now.
Seven-loop card
Binds in principle and cannot bind yet, because its checker, a symbolic engine able to do the seven-loop calculation, does not exist yet.
Proven
The corpus of eight talks is sealed and verified MATCH. The judge's refusal on all six convergence claims is on the record.
Open (UNVERIFIABLE)
The talk-to-design convergence mappings (no oracle). The neutrino card (data not yet in hand). The method-template card. The seven-loop card (no engine in reach).
What would prove this wrong
If a future card gets in without a named falsifier, the intake check has been breached, and the forge has stopped working as a forge.
Study 3 · The C3 result: what a bound claim looks like
C3 is the central case of this thesis. It is the one place either forge crossed from bid to bound, and it shows where the boundary falls: the verdict binds on one side of the line and stays a bid on the other. The full experiment has its own page.
The claim comes from a research talk. A thermodynamic chip is built so that the chip itself is a stochastic differential equation, a rule for how a noisy system moves over time. As its noise settles, in the speaker's words, the chip behaves "sort of" according to the inverse of a matrix. Put plainly, its settled state gives the answer to a standard algebra problem. The simulate leg of this claim can be tested now, with no outside party, because linear algebra authors the criterion. The process being simulated has no hand in it.
The method is re-derivable. Simulate an Ornstein-Uhlenbeck process, a standard model of a noisy system that is pulled toward an average and keeps jittering around it. Run it for a symmetric positive-definite matrix A, a well-behaved kind of matrix, of size n=4 with condition number 2.33, a measure of how sensitive the problem is to small errors. The run uses 6000 chains, meaning 6000 independent simulated runs, and 40000 steps. The settled average, called the stationary mean, should equal the linear solve: the exact answer to a set of simultaneous equations, which ordinary math software computes. The settled spread, called the stationary covariance, should equal the matrix inverse. A standard numerical solver supplies the reference values, and the record names the reproduction script. Against a principled five-percent limit, the measured results were:
Stationary mean against the linear solve: relative error 0.99 percent, well inside the five-percent limit. MATCH.
Stationary covariance against the matrix inverse: relative error 3.3 percent, inside the limit. MATCH.
Both simulated results landed inside the five-percent limit, so the math check passed.
How to read this: each bar shows how far the simulation's answer sat from the exact answer, as a percent error. A shorter bar is better. A bar that ends left of the dashed line is inside the limit and passes.
This checks the math in simulation only. No physical chip was measured.
The perceived shape is the simulated stationary statistics. The criterion, the exact inverse from a standard solver, came from outside the checker. Anyone holding the script, the seeds (the starting numbers that make a random run repeatable), the seals and the measured numbers can re-check the proof. All three reconcile steps are present, so the MATCH binds. The 0.99 percent mean error against a five-percent limit is a number anyone can reproduce, and it could have come out the other way.
The record itself states the boundary: this confirms the simulate leg only. The mathematical claim, that the process settles to the inverse, is now a re-checkable MATCH. Whether a physical thermodynamic chip realizes this within its noise and device limits is the render leg, and it remains UNVERIFIABLE here. The mathematical claim is bound. The physical-device claim is an unwitnessed bid, because its criterion, measurements from a real chip under real noise, is not in hand. One claim has two legs, and the verdict splits where the external criterion stops being available.
Proven
The simulate leg of C3, as a re-checkable MATCH. The measured relative errors were 0.99 percent for the mean and 3.3 percent for the covariance, against a five-percent limit, reproducible with the named script and sealed. The relative errors are the independently re-derivable quantity. The judge's margins are the record's reported figures, and this page has not re-derived them (see How we know).
Open (UNVERIFIABLE)
The render leg. No measurement here shows whether a physical thermodynamic chip realizes the mechanism within device limits. The record marks it UNVERIFIABLE, and so do I.
What would prove this wrong
Re-run the script against a fresh symmetric positive-definite matrix. If the stationary mean and covariance fail to recover the linear solve and the inverse within tolerance, the simulate MATCH has drifted and must be downgraded to DRIFT. Anyone who cites C3 as evidence that a physical chip works has crossed the line between the simulate leg and the render leg, and overclaims.
Study 4 · The research diet: the same boundary, found independently in physics
The research diet was a reading program. It read eight talk transcripts, sealed them, and mapped them to the system's design, with every mapping marked interpretive. This study sets the mappings aside, since the judge refused them, as Study 2 noted. Several talks, independently, show the boundary between a bid and a bound claim inside their own fields. Two readings are faithful to their source and worth keeping.
The experimental turn in a foundations dispute. One landmark result turned an interpretive dispute into an experiment. A class of hidden-variable theories, in which properties set in advance decide what a quantum measurement shows, implies that certain correlations have a limit. The competing theory predicts stronger violations that can be measured. Over decades, the experimental arc closed loopholes one at a time. This is the bid-to-bound transition in physics: a claim that was an unwitnessed bid became a bound verdict once someone built a criterion that could fail and could be repeated. Using this result as a pattern for the system is still UNVERIFIABLE, because that use has not itself been run through any measurement. The physics is settled. The analogy to the system is a labeled interpretation.
"We do not know the rules of the game." One talk pairs a falsifiable, dated prediction with repeated caveats, in the speaker's words, that "we do not know the rules of the game." The diet reads this caveat as the working scientist's version of UNVERIFIABLE. The scientist states a dated falsifier and, in the same breath, names the part he cannot yet ground. The forges do the same thing: they bind where the criterion exists and mark UNVERIFIABLE where it does not.
The diet's cross-talk synthesis draws four habits from several talks at once:
Carry uncertainty in the open.
Count only a full pass as a pass.
Map auto-formalization, which turns a written argument into a form a computer can check, and prove-or-disprove onto MATCH, DRIFT and UNVERIFIABLE.
Turn a dispute into a re-checkable measurement, then close the loopholes.
These habits come from careful physics, and the forges try to run them on the system's own claims.
One caution carries over from the diet. The cross-talk mappings are the author's interpretation, and the judge declined to certify them. So this study cites the talks as faithful readings. It makes no claim that physics verifies the design. The transcripts are sealed and can be re-checked: ten items verified MATCH, the eight talks plus the run's metadata. The interpretation carries its label.
Proven
The eight transcripts are sealed and verified MATCH. The physics quoted is faithful.
Open (UNVERIFIABLE)
Every talk-to-design mapping. The judge refused all six, and no oracle exists to measure them.
What would prove this wrong
Re-run the seal against the corpus. If the ten-item MATCH count does not hold, the provenance claim is refuted. Anyone who cites this study as proof that the talks verify the design has misread it against the record's own warning.
Study 5 · The viable-visualization thesis: the same boundary, made quantitative
The viable-visualization thesis asks whether a finite system can show mathematics to a person faithfully, within a scope it declares in advance. It is here because it is the one place where the boundary of this thesis gets a measure. Its central object is coverage under an explicit scope. A finite perceiver has finite variety, meaning a limited number of distinct states it can tell apart. Mathematics has no such limit, so no finite visualizer can cover all of mathematics. The system is incomplete over all of mathematics and viable over any bounded class it declares. When it cannot meet the bound, it fails closed: it withholds the pass and emits the same UNVERIFIABLE that the reconcile already emits.
That is Claim 3 made quantitative. Here the line between bound and bid is whether the channel, the picture that carries the mathematics to the viewer, holds enough variety to recover the criterion-invariant, the property the check looks for. The thesis ran sixteen experiments that attacked its own clauses on purpose. Two of them show the same habit of self-correction:
One experiment downgraded the thesis's own claim of one conservation law, a single quantity that always stays the same, to three mechanisms with one necessary variety bound, a minimum amount of variety a faithful picture must carry. My record of that experiment says it downgraded my own overclaim: proof before trust, applied to the proof.
The live reconcile loop reached a certified majority of its cases and left a residual: a set of genuine variety deficits, cases where the picture carries too little variety. The record reads that residual as the variety bound in operation, which is expected behavior.
I include this thesis with one marker the others do not need. Its experiments run against the project's own substrate, meaning its own software and data. Its MATCH and UNVERIFIABLE verdicts are therefore internal to that substrate. C3's linear solve and the survey bound are criteria authored entirely outside the project, and these verdicts have no such outside criterion.
The source is also uneven about its own grand unification, a proposed single structure meant to tie its separate findings together. I report both readings, and I do not pick the flattering one. On its proposed unifying structure, the document points two ways. One passage still marks the structure as pure theory that its simulations cannot settle, and lists it among what remains open. A later passage in the same document claims the structure resolved and formalized, and declares the arc complete. That is an internal inconsistency in the source.
Given that tension, this wrap-up takes the conservative reading, and I flag it as a deliberate underclaim. I do not endorse the unification as a proven result here, for two reasons. The same document says its own simulations structurally cannot settle it. The claimed proof is also internal to the project's substrate, with no check against an outside criterion. I also record that the source reaches a stronger, resolved verdict elsewhere, and I choose not to carry that verdict forward as bound. A reader who re-checks that passage and judges the formalization sound may upgrade it; until then I hold it as conjecture.
Proven (within its substrate)
The master clause, the thesis's central statement, survived adversarial attack across sixteen experiments, and two of the thesis's own overclaims were caught and corrected. The experiments led the project to build new working parts.
Open or conjectural (UNVERIFIABLE)
The cross-domain variety measure. The thesis concludes that no single substrate-free measure exists, which is itself a result, and it offers no positive measure. The unification structure, held here as conjecture, with the disclosure that the source claims it resolved elsewhere.
What would prove this wrong
A bounded class where the variety bound is met and yet no rendering is comprehensible. A class where two perceivers cannot reach consistent verdicts against any external criterion. Compositional fidelity, a picture staying faithful when its parts are combined, failing beyond repair for a target domain. For my conservative reading: if an independent party re-derives the unification against a criterion authored outside the project's substrate, my conjecture label is refuted and the source's resolved verdict stands.
Why the boundary is the honest shape of verification
Put the studies side by side and they show one pattern. A verdict binds where its criterion is externally witnessed and re-derivable, and it reverts to a bid everywhere else.
Every study splits at the same line: the left side has an outside check behind it, now or in principle, and the right side has none yet.
How to read this: each row is one study. The left box is the part that holds, or could hold, because an outside check backs it. The right box is the part no outside check backs yet. The solid line between them is the boundary this page is about.
The list below gives the same six rows in words.
Learning Forge
Binds on corpus provenance, since the seals re-derive. Stays a bid on all six claim cards, since no judge has run.
Discovery Forge, neutrino card
Binds when the dated survey data arrives. Stays a bid now, by construction, since the data is not in hand.
Discovery Forge, C3 result
Binds on the simulate leg, where the linear solve is the criterion. Stays a bid on the render leg, since no physical chip was measured.
Discovery Forge, seven-loop card
Binds in principle, since each loop order is checkable. Stays a bid now, since no symbolic engine is in reach.
Research diet
Binds on transcript provenance and on the physics itself. Stays a bid on every talk-to-design mapping, since no oracle exists.
Viable-visualization thesis
Binds on variety recovery within its substrate. Stays a bid on the cross-domain measure and the unification structure.
Every row shows the same boundary, and the studies are built to report it. A verification system that claimed to bind everywhere would be claiming aseity for its verdicts: its MATCH would stand on its own footing and hold in every domain with no external criterion. A self-standing verdict is as incoherent as a self-standing existence. The empty Learning Forge slots, the Discovery Forge's UNVERIFIABLE convergence claims and the C3 split between legs all sit at this limit. Each one marks where an outside check stops today, and the checker cannot move that line by trying harder.
A verdict that completes all three reconcile steps binds, and it binds only in the domain where all three were available. The C3 simulate result completed all three, and it binds. The C3 render claim cannot complete the second step, since no external device measurement exists, so it does not bind, and the record says UNVERIFIABLE. No fourth step lets a verdict bind without an external criterion, because nothing has aseity to appeal to.
What this thesis proves, and what it leaves open
Proven and bound (evidence cited in the source records):
The Learning Forge corpus is sealed and re-derivable. Its six claim cards are grounded in hashed sources, and all six verdict slots are empty. Of ten modules, five are grounded, two are partial and three are ungrounded.
The Discovery Forge admits only cards with a named falsifier. It carries three anchored falsifiers: neutrino mass, the thermodynamic process, and seven-loop quadratic gravity. The judge's refusal on all six convergence claims is sealed.
The C3 simulate leg is a real, re-checkable MATCH: 0.99 percent mean error and 3.3 percent covariance error against a five-percent limit, reproducible with the named script and sealed. It is the worked example of a bound verdict. The recorded judge margins are the record's reported figures, and this page has not re-derived them independently.
The research-diet transcripts are sealed and verified MATCH. The physics quoted is faithful.
Left open (UNVERIFIABLE):
All six Learning Forge claim cards. No judge has adjudicated them.
The Discovery Forge's neutrino card: undecidable until the data for the three-to-four-year survey window is in hand.
The C3 render leg: no physical-chip measurement exists here. The math binds, and the device claim stays a bid.
The seven-loop card: out of current reach. No engine can carry a seven-loop perturbation calculation.
The method-template card, and every research-diet talk-to-design mapping: interpretive, with no oracle, and the judge refused them.
The viable-visualization thesis's cross-domain variety measure and its unification structure: held here as the thesis's own conjecture, with the disclosure that the same document claims the structure resolved elsewhere. This page does not endorse either as proven.
This thesis as a bid about itself. By Claim 1, this wrap-up is a made thing with no aseity, and it was written to show that. Its three claims bind for a reader who inhabits the practical standpoint of caring whether a verdict is grounded. A reader who withdraws from that standpoint can decline them. Its evidence is the record paths and seals cited in the underlying draft, which is not published. If those records are altered, or the seals fail to re-derive, the matching claims here revert to bids. The thesis earns whatever bindingness it has the same way the forges do: it names the external criterion it was checked against and invites the re-check.
How this page was written
This page stays faithful to its sources. Every section states what is proven and what is UNVERIFIABLE, with a falsification condition or a citation to a source record for each non-trivial claim. The four forge and diet records are quoted with anchors in the underlying draft, so a reader of that draft can re-check the reading against the source. The viable-visualization thesis carries a marker that its verdicts are internal to its own substrate. Its larger unification is reported with both of the source's competing self-verdicts and held here as conjecture, so this wrap-up does not overclaim that it rests on a settled structure. The bid-versus-bound structure comes straight from two earlier works, applied to verdicts: the conferred-existence thesis's own figure of ground and conferral, and the reconcile thesis's own perceive, check and carry operation.
A verdict has the same ontology as everything else in the conferred-existence thesis. It is conferred, relational and re-spoken. It binds while its criterion is witnessed and re-derivable, and it becomes an unwitnessed bid the moment you step outside that domain. The forges already behave this way, and C3 shows a verdict that crossed, on one leg, into a bound claim.
How we know
Does not prove
This page does not show that a physical thermodynamic chip works, that physics verifies the system's design, or that the viable-visualization unification is settled. Each stays UNVERIFIABLE above.
Where the evidence lives
The record citations and their anchors sit in the underlying draft and are not linked from this page. The specific seal hashes and the absolute record paths stay there too.
The C3 margins
The two relative-error numbers and the five-percent limit are the quantities that re-derive reliably: re-running the script reproduces them directly. The recorded judge margins, 0.802 and 0.340, come from a single record line. They are the weakest evidence link in this thesis, because the record names the script and does not show the margin computation itself. Treat the margins as the record's reported numbers. This page has not re-derived them independently.
The Learning Forge seals
The corpus has a digest seal, a run seal, and stored items that can each be retrieved by their SHA-256 hash. The hash in a card is the retrieval key, so any reader can re-derive the inventory.
Check it yourself
The C3 page sets out the matrix, the method and the seed-fixed script behind the 0.99 and 3.3 percent errors, so a reader can follow the simulation and repeat it.
Status
Reviewable draft, dated 30 June 2026.
Author and attribution
Author of record: Zain Dana Harper. The attribution trail is git history, the cited record paths, and the corpus seals quoted from those records.
Scope
Public-safe: no secrets, and no client or private data.
A wrap-up thesis tying four recent studies, plus one added case, to the reconcile thesis. Every claim above is paired with a falsification condition or a citation to a source record. Back to Research.