What Does Not Derive It

This section is the register of every road to α's value that has been walked inside the corpus, with the locus of each and the reason it closes, so that no road is walked twice. One rule governed every road: nature decides, not arithmetic. No road was allowed to end in a search over functions of the counts. Where a small set of candidates was tried and reported negative, the set is named so that it stays closed.

One thing is not a road and is recorded here so no one enters it as one. The electron’s three length scales stand in adjacent ratios of α (§2, §4). That is definitional — the classical radius is built from α — and no ratio among them can return α as a result.

The roads, twelve closed

One. The partition. AP42 builds the dark partition from the six, the twenty-one, and an exponential of their ratio. The wall's verification code has no α in it. Reading α off the partition is therefore not circular, and there is no α there to read. And the reason it cannot be read there is not dimensional — six over twenty-one is dimensionless and so is α. It is that nothing in the corpus maps the partition’s occupancy ratio onto a window’s transparency. The two are dimensionless quantities about different things, and no derived bridge joins them. Closed.

Two. The count. AP45 §6 rules it inside the family already: the channel counts are geometry, fenced from arithmetic, generating no fraction, and connected to nothing they do not name. That is a locked ruling and it is the primary citation here. Beyond it: every pure number the corpus produces is a count or a shape, and one hundred and thirty-seven is neither. KS-30.4's between-family search found 2,971 five-leaf formulas at the proton's accuracy; this paper adds α to no such list. The four invariants of the event graph and the held graph — the square roots of twenty-one and eighteen, and the line-graph radii eight and seven — were tried and none is near α; that set is closed and no further invariant will be tried. Edge-on there are no counts (§7). Closed by rule.

Three. The cycle turned round. AP46 §9 gives one cycle as (21/18)·α⁻¹⁸ ticks. Read backward on the measured age it gives α to about 80 ppm, sitting 173 ppm above the measured value; the proton row needs α to about 0.2 ppm. Circular — the corpus gets the age from α — and four hundred times too coarse if it were not. Closed.

Four. Gravity read back. Gmₑ²/ℏc is AP28's object and AP44's realised formula; The Lock's G = 2κ/mₑ² is the same equation by the bridge 2κ = α²¹(1+1/π)ℏc/(1+α), and the bridge is tautological in AP08's own words (§5.1: the coefficient has been relabelled, not independently derived; KS-I.3, KS-I.4). Read backward on measured G it gives α to about 1 ppm, 17 ppm above the measured value — the G row's −16σ wearing α's units. AP24 §3.3 says the same in its own words: the holding-limit form and the channel-count form are one equation. Taking G as the one measured input is legal and buys nothing. Closed.

Five. Gravity against the clock. Eliminate α between AP44 and AP46 and a relation remains between G and the age with no α in it, exponent 21/18. It is the two rows multiplied: fed the measured age it returns G 0.33 per cent high; fed measured G it returns the age 0.28 per cent high, which sits at +1.9σ on the Planck bar while the age row itself remains at its own +2.1σ — the elimination adds no test, so the row is unchanged. It creates no test the two rows did not already carry, and it loses the attribution they keep. KS-CCC.3 and KS-STRETCH.3 own it; it is not a switch of its own. It is a large-number relation with a structural exponent; T is the cycle, a constant, so G does not drift. Closed as a road; recorded as a relation, with one property the separate rows do not have — it holds whatever the coupling is, so it tests the common rate without assuming the rate is the electromagnetic one, which is D-EM.1's question asked from the side.

Six. The holder recursion. AP49's chain, C = 21α + (16α/1836)·C, prices a unique total cost for every coupling and prefers none; the closed form reproduces the proton at +0.59σ for α as measured and a different proton for any other α. Closed.

Seven. The drip. Let each opportunity admit one grain or none, w the load of one grain, ℓ the maintenance per opportunity. A bounded store forces the admission fraction ℓ/w — too much and the load climbs, too little and it drains, the tipping picture in one line. It selects the fraction once ℓ and w are fixed by different things. They are not: in the corpus both carry the coupling, so the condition reduces to an identity in it. The author rules the flow-quantum one whole grain per now, read from the side, at the same rate at every window. On that ruling the admission fraction is fixed by the rope’s length in grains, which is the bracket of the axiom and not something the drip determines. Closed on the author’s word and on the identity together.

Eight. The floor. AP06 §7.2 derives that the leak is positive, because finite c forbids perfect absorption, and identifies the value with the measured leakage, and says so. A field calculation of readability — the least error in telling two charged sources apart from their field record — confirms it: with unbounded repetition every positive coupling reaches any tolerance. A threshold above zero needs a resource rule and a tolerance; neither is in the corpus, and a resource rule taken from the hold would carry the coupling in it and select nothing. AP12 §1.3 is the same shape for ℏ: minimality pins the unit, not the value. Closed at zero.

Nine. The fixed point. AP24 §4.3 names ε = f(ε). The phrase means the self-consistency fixed point of the six readings here, and not the renormalisation-group fixed point of §3; the two senses are unrelated and the paper never mixes them. The six faces of the residual carry two pairings, both locked, and they are not rival readings of one relation — they are two relations, each derived for its own work. AP10 §5.1 pairs by axis: G with c, the substrate stiffness with α_em, mₑ with t, the residual's cross-check on 3+1. AP45 §6 pairs by side: mₑ with G, α with c, the substrate with τ, the balance partition that gives ten and ten plus the one. Composed, the two close the six into two three-cycles — G to α to t, and c to mₑ to the substrate. Every arrow in both is a translation between readings of one object, so the map composed around each cycle returns its input, and every coupling is a fixed point; AP24 §3.1 calls the relations tautologies and is right. An explicit family shows it: for any q define ε₀(q), κ(q), G(q) from q and every displayed relation holds. One precision, because the closure is narrower than it looks: the permutation closing is combinatorics about labels, and a permutation of labels is not yet a quantitative map between dimensional readings. The cycles compose to the identity because AP24 §3.1 makes every arrow a translation, and if any arrow instead carried an independently derived normalisation the composition would not be the identity and a numerical fixed point could arise. So the road is closed at the relational level and no further, and it reopens the day one arrow is a law rather than a dictionary entry.

Ten. The earlier thread. From the author's private notes. τ × t_H × α = 0.999, a tautology. V × t_H = 1, the construction. (V+D)/α = 2t_H plus 1.6 per cent with no source. A race between c and G giving 137, off by forty orders. 1/α ≈ t_H + τ(t_H−1)(1−e^(−t_H/τ)) at 137.376, the form chosen after it landed. The same with −1/3 at 2 ppm, the 1/3 having three sources and therefore none. And α = cycle + black-hole leakage → expansion, closing on magnitude (the horizon leakage is 10⁻¹⁷, AP06 §1), on category, and on α's absence from the partition. All withdrawn on the prior-derivation test. Closed.

Eleven. The material face. Two stiffnesses whose ratio is c² admit two independent combinations; the electromagnetic vacuum reads the second as an impedance, and α is that reading in the grain's units. The impedance reading cannot be turned into a measurement route: since the 2019 redefinition the elementary charge and ℏ are exact, so the resistance quantum is exact and the vacuum impedance is known only as well as α is known, being proportional to it. Reading α out of it returns what was put in. Asked whether the picture reads a second thing from the two stiffnesses, the author answered with the axiom's balance and its one-grain residual, and named no second reading. Closed as an empirical inversion, by the definitions, and closed on the author’s word that the picture names no second reading. Not closed as a first-principles problem: deriving the substrate’s normalisation from S, B, R and C without α is not this route — it is the missing second scale itself, and this paper does not have it. Road Nine’s missing edge-law and this road’s missing normalisation are one object seen from two sides.

Why five of the roads close alike

The roads close for their own reasons. Five close alike, and the likeness is worth stating — but it is narrower than an earlier draft of this section claimed, and the narrowing is on the record rather than quietly done.

Withdrawn. An earlier draft argued that an equation in the coupling whose coefficients are the picture’s counts cannot have a root at α, on a standard bound on the roots of a polynomial. The argument required every coefficient to be of order one, and the same paragraph listed eighteen hundred and thirty-six among the coefficients. With that admitted the bound’s lower edge falls to about one in eighteen hundred, and α sits well above it: the bound does not exclude α. Nor is the class thin. Quadratics whose coefficients are products and ratios of the picture’s own counts produce roots at α’s value in quantity — thousands within one per cent, dozens within a hundred parts per million, two within one. The sentence “that class has no member with a root where the value is” was false, and it is withdrawn.

What survives, and it is the spine. The five roads as actually built were made from single counts and simple shapes, and not one of them landed. The closest any of them came to carrying a coefficient of α’s size is the holder recursion’s sixteen over eighteen hundred and thirty-six, which is α times 1.194 — nineteen per cent away, and it did not land. The others used coefficients that are not α either: one over eighty-four pi is α times 0.519, of α’s order and not at it, and six over twenty-one is α times thirty-nine. What the graph invariants actually returned was a fifth, a quarter, an eighth and a seventh — used and returned are different things and this paragraph had them interchangeable. Two of the five reduce to identities in the coupling and return no root whatever. The five are the tipping condition, the holder recursion, the readability threshold, the fixed point and the graph invariants — named here because the Conditionality Footer’s firing condition depends on which five they are. That is one failure, five times, and the same one. It is a statement about what was built. It is not a theorem about what could be built, and this paper no longer offers one.

The abundance that broke the theorem is a finding in the other direction, and it belongs to §11.

Twelve. The ledgers. Read per tick, the cut costs α-worth and the hold leaks α to the eighteenth; both grow cheaper as the coupling shrinks, so nothing pushes back and there is no balance to strike. Read per cycle, one window makes α to the minus eighteenth cuts at α-worth each and something does push back — but setting that spend against one grain shed per cycle gives an equation whose root is at α only if a structural number of about two times ten to the thirty-sixth exists to put on the other side. The picture produces no count within thirty orders of magnitude of it. Closed on the size, which is the whole of the argument and needs nothing else. Closed for this ledger construction and for the corpus’s existing structural numbers: an independently derived scale of that order would be a new road, not a rescue of this one. Which ledger the break is kept small in therefore does not bear on the value, and the question is answered by being made irrelevant.

The two roads that are open

Thirteen. The running. Two things have been called one road here, and they must be separated before either is judged. AP24 §5.2 names Path 1: start where the coupling is of order one, run it down with the charged spectrum, land. AP24 calls the path rigorous and calls the spectrum a Notebook V programme.

The toy route is closed. Its boundary condition is a coupling of order one at the Planck scale, and ordinary Standard Model running does not produce one there, which is why §0.8 refuses that normalisation. The size of the shortfall is a diagnostic and worth recording, so long as it is not read as a prediction: in the toy’s single-coupling convention, forcing that boundary to land on α at the electron would require an effective coefficient near 16.6 against about 10.7 from the charged fermions counted the same crude way. That gap is evidence against the toy’s boundary. It is not a measurement of missing particle content, because the coefficient is not a count of charged stuff — spins, representations, charged vector bosons pulling the other way, thresholds and the electroweak mixing all enter it, and above the weak scale the honest treatment runs the electroweak couplings rather than one electromagnetic coefficient. AP24 §5.6 said the toy was not the full function; this is the arithmetic of how far it is from being it.

The general route is open. Derive an ultraviolet boundary value from S, B, R and C. Derive the charged spectrum. Let the flow carry the one to the other. That route is listed here so that no one reading §10 alone concludes it was closed. It is the only direction the corpus has ever looked in that is neither a count nor a power of the coupling against another power of it. AP24 §5.3 points at it from the same side, because it points at the renormalisation-group flow as the place a non-algebraic value can arise, and the logarithmic scale dependence lives in that flow’s integrated solution. D-EM.1 owes what the running means for the corpus’s α. Three pointers, one road, open and owned elsewhere.

And the price is exact about what is owed. What the flow supplies is the map. It does not supply the ultraviolet boundary value and it does not supply the spectrum, and those two are the whole of the outstanding derivational content. The road is open at the machinery and shut at both ends. It is falsifiable at the spectrum before it is ever falsifiable at the number, and it is not this paper’s.

Fourteen. The ripple. The author’s describing of 12 September 2026, with his ruling on it the same day. The arena is not still. Gravity deforms the substrate and the substrate resists, and the deformation that opened at the break never settled — it settled into a beat. The beat is what the landings sum to — every landing one beat, the rate of the beats the tick — so there is no phase of it that any site reads. The floor of the arena ripples like the skin of a drum. A held configuration is therefore maintained on a moving floor, and maintaining it there costs. The same ripple is what carries a window through its opening and its closing: without it the window would settle open or settle shut, no next grain would pass, and the tick would stop.

The ruling, and it is what holds every frozen row in place. The ripple is the cause of a cost the corpus has already priced. It is not a new cost. AP47 and AP49 booked it; this says why it was there to book. No number moves anywhere, the pre-registration stands as written, and KS-HOLD.1 is untouched.

What it supplies, and it is real. AP49’s chain says every holder must itself be held and each costs a share of the last. It does not say why the first holder costs at all. The ripple answers that: because the floor beneath it is moving. It also says why a static census would miss such a cost at all. A symmetric vibration averages to zero at first order and costs at second order in its own displacement, so a census that counts the mean configuration correctly misses the variance around it entirely. That is the shape of the omission the neutron’s firing exposed, and the corpus had the bookkeeping for it before it had a reason. It is not the same thing as putting the term at second order in the coupling, and the paragraph below says why not.

It also gives a reason for where the corpus’s costs sit that does not depend on the shape of the wave. Any smooth crest with non-vanishing curvature, touching a threshold, opens a window whose width goes as the square root of the excess above it; a flatter crest scales otherwise. So a first-order window needs only a second-order excess, whatever the waveform. The corpus’s own pattern — a first-order coupling with second-order costs standing behind it — has that geometry underneath it. It is a rhyme and not a derivation: it says why such a split is natural, not that the corpus’s particular second-order terms are that excess.

And the road repairs the worst problem in its own first describing. A rope read as one slow wave a hundred and thirty-seven grains long would have to travel at a hundred and thirty-seven times the speed of light. Read instead as the coherence of the hold — the carrier staying at the electron’s tick while the held configuration keeps its phase history for one over the coupling of them — the long scale appears with nothing exceeding c. That reading is conditional on the loss per tick being the coupling itself, so it accounts for the ladder’s structure without supplying its size.

Why it is not a road to the value, and this is shown rather than argued. Write the ripple as a substrate field carrying an overall response normalisation in front of its Lagrangian, with a grain coupling to it at some strength. The free equation of motion does not contain that normalisation at all. Normalise the field canonically and exactly one quantity moves: the grain’s coupling divided by the square root of the substrate’s response. So the free dynamics — the propagation speed, the free frequency, the phase structure — are unchanged across a continuous family of couplings.

Two things are not, and the paragraph is narrower for saying so. The amplitude of the ripple, once grains are sourcing it, moves with the normalisation, and a hold cost depends on the amplitude. So does a threshold gate’s open fraction, which is what the coupling is: change the amplitude and the window opens for longer, which is a moved closing time. The no-go is therefore about the free dynamics and not about the whole picture. The free part is blind to the parameter; the gate and the costs are not. A describing of the free part can be entirely right and settle nothing about the value.

The one-record-is-one-quantum-of-action condition does not close it either. That fixes the unit. The matrix element through which a grain actually couples carries the same square root, so the fraction of the unit involved in one exchange stays free. And the obvious dynamical closure fails on its own arithmetic: a driven ripple in steady state delivers exactly as much power as it loses, for every value of the loss coefficient. Stable cycling says the universe can keep the ripple running. It says nothing about how strongly one grain couples to it. That is Road Seven’s identity again, in other clothes.

Found twice, by different mathematics. A second construction reaches the same obstruction from the discrete side, and it starts from less. Require only that the one-cycle return map have no preferred phase origin — which is S, not an assumption — and the map is forced to depend on the amplitude alone. Take its lowest non-trivial order and the still arena is an unstable fixed point while a finite ripple is a stable one, which is the describing’s own claim in dynamical form rather than an extra premise. Now count. The map has four parameters: the drive, the saturation, the window’s response and the phase advance. Rescaling the amplitude is a change of coordinate and nothing else, and it removes exactly one of the four. Three survive it, and the coupling is a product of two of them. Nothing in the construction fixes either. An earlier draft of this paragraph said that exactly one combination survives, which is false — the drive survives on its own, and so does the phase advance — and the correction is printed rather than made quietly. What is true is the count, and the count is the no-go: the coordinate freedom is one parameter’s worth and the unknown is more than that. The field calculation reaches the same wall by a different road, which is worth something; the two constructions are not the same theorem and this paper no longer says they are.

Two smaller reasons stand behind that one. The drive is the coupling: with any loss at all an undriven ripple decays, so the ripple must be driven, and what the potential side delivers to one window in one tick is the cut fraction. And the variance fixes an order and not a value: the cost goes as the square of the ripple’s amplitude, so for it to land at second order in the coupling the amplitude must itself be of the coupling’s order, and nothing in the picture supplies that.

What the road does supply. The missing quantity now has a physical name instead of a placeholder. It is the response of the moving substrate to one grain — how hard the floor is to couple through at the actualisation frequency. Written the ordinary rationalised way, the coupling is the square of that response over four pi, which is to say the question has become what fixes the charge. Physics leaves it in the same place and says so. That is not progress towards the value and the paper does not present it as such. It is the first entry in this register that names the missing second scale rather than only saying what it is not.

Four closures found inside this road, recorded so none is walked twice. Energy balance selects nothing: if one landed grain replenishes the ripple that released it, the coupling cancels from both sides and a ratio of inertias is all that is left. A complete coherent flip per tick is not it: a full two-state swap in one tick needs a coupling of π over two, which is of order one and not of α’s. A driven passage through an alignment with a latch selects nothing either: the outcome depends on the coupling squared over the sweep rate, so any coupling whatever is reproduced by choosing the sweep. And the tempting mechanical reading of the axiom line — the excess over exact balance read as the break, the release fraction read as the coupling — forces the coupling to two over π squared, about a fifth. Dead on its own arithmetic.

The calibration pass, and why it is not the test. The ripple was run against the proton’s second-order term and the neutron’s flip cost, and it reproduces the form of both. It does so using the corpus’s own counts — the twenty-one, the sixteen, the eighteen hundred and thirty-six, the two orientations, the four dimensions — every one of which was fixed by AP30, AP47 and AP49 before the ripple existed. Recovering a form from the numbers that fixed the form is a consistency check, and consistency is worth having. It is not the test. The test this road must pass is the one stated below: those corrections without their coefficients.

Two fences, recorded so the next desk does not build over them. A loss going as the square of the motion leaks more when the ripple is larger, while AP44 and AP45 shed one ε per now, flat, whatever the crowd; those are different laws and the discrete one is the locked one. And a drum head that rises and falls as one is a universal clock, which AP52 has live switches against — the picture survives only in the author’s own terms, that the vibration is the now, shared by every window and read by none, while the ripples a site can actually read travel at one grain per tick, which is c. No fork is needed there, and an earlier draft opened one in error: §3 already rules that a trillion windows firing at one now is one grain edge-on, so unison puts nothing extra through the neck. The ruling is the first sentence and no more.

Open, with its price and the order of the work stated. The ripple is a mechanism, and the corpus has no other account of why a hold costs at all, so the road is not closed. Before it is ever asked about α it must recover the proton and the neutron correction blind, from the ripple alone, without their measured coefficients. That is its own paper and its own switch. If it cannot pay that price it explains the picture and derives nothing, which is an honest outcome and is stated here in advance rather than discovered afterwards.

The price was attempted on 13 September 2026, and the road did not pay it. A construction was built from the lodged describings, frozen and hashed before any comparison, its digest carried with it and recorded on receipt by a desk that had not built it, and compared there. What the instruments page timestamps is the record of when each file was lodged, and this paper asserts nothing beyond what that page shows. The construction reproduced the form of both rows. Every count in it, however, coincides with a decomposition the corpus had already locked, and the writing desk disclosed in advance that it had spent two days with those decompositions. A form recovered from the numbers that fixed the form is what the calibration pass already was.

And the comparing desk found the steering, in the place the construction itself nominated. The construction's one argument that was not already in a locked paper was that time enters the support as a fourth axis, because a re-establishment is an event and events happen in time. That is precisely the fourth axis the corpus's own decompositions already carry. Worse, it was applied only where it lands: a first hold is also an event, and the corpus prices a first hold without that fourth axis, which the construction avoided by declining to derive the first-order term at all. Disclosed as non-blind, and still steered.

The one new claim in it failed. The construction predicted that a particle's escape support and its re-establishment support are different numbers. That prediction holds only under a reading of the leakage geometry the corpus does not use, and under the reading it does use the two supports coincide at the neutron. The construction also reached for a full solid angle, where the corpus's own geometry is the grain over the disc of its aperture. Dead at the neutron, on the corpus's own reading.

What survives is the cause and not the coefficients. A hold on a moving floor is re-established rather than maintained; a re-establishment is a writing; writings cost; and the cost compounds because each re-establishment must itself persist. That is worth a dated note beside the two papers that book those costs without explaining them. It is not a recovery of anything, and this paper does not present it as one.

Two ways remain to pay the price, and both are named rather than promised. A genuinely blind formalisation, by an instance with no corpus and no paper numbers in front of it, under the sealed-envelope discipline the corpus already uses; every desk that has worked on this road is now disqualified from that route. Or a prediction about a configuration whose coefficients are not on the wall, the deuteron being the obvious one. Until one of those is done, Road Fourteen is a mechanism and not a route, and the register carries it as a miss published rather than as a road still being walked.

The two open roads are open in different senses, and the difference should not be left for a reader to work out. Thirteen is open in the strong sense: pay its price — a boundary value and a spectrum — and a value follows. Fourteen is open in a weaker one: pay its price and what follows is the missing scale with a physical name and two candidate identifications — AP15’s KS-31, the conserved balance between wave-side coherence and field-side connection, and AP43’s owed normalisation of its lean coefficient — not the value. Its own result proves the picture is blind to exactly that parameter. It is on the register because it converts a placeholder into a named object, which is progress of a kind, and not because walking it reaches the number.

One correction carried from the describing, and it comes from C. The ripple is one grain per tick and a rope is many ripples. A ripple one rope long per tick would travel at a hundred and thirty-seven times the speed of light. So the coupling stays a ratio of two rates — the tick against the completion — and not a count of anything.

What a live route would have to contain

Counting does not land on α. It reaches α’s order easily enough — sixteen over eighteen hundred and thirty-six is α times 1.194 — and the register records why none of those approaches lands. What counting does not supply is a dependence on scale. One family of functions does, through its integrated solution, and it is the exponential and its inverse. The corpus has exactly one exponential, in AP42’s partition, and Road One records that it contains no α and is spent. Road Thirteen is the other place such a function appears, and there it is the running.

The corpus said this before the hunt began, and said which transcendental. AP24 §5.3’s necessity claim about irrationality is withdrawn by the dated note of §13; what stands is the weaker observation it also makes, that a fixed point of a map built from the renormalisation-group flow is generically non-algebraic. The flow law itself carries no logarithm; its integrated solution is where the logarithmic scale dependence appears. Road Thirteen is that same door named as a route. It is a convergence and not a plan, and it names the debt, because the beta function is the running and the running is D-EM.1.

This is not an instruction to look for another. A form chosen because its exponent lands is the fit of §11 wearing a better disguise than most, and the register of §11 exists for exactly that case. An exponent would have to be argued from S, B, R and C, written down, and lodged before the value is consulted. Under the protocol, and not otherwise.

It is recorded for a second reason, which belongs to physics and not to the corpus. Renormalisation-group evolution is an established mechanism by which a dimensionless coupling at one scale fixes its value at another, through logarithmic dependence on scale. That is the kind of non-algebraic map the counting routes lack. It is not a claim that α is of order one high up and driven small by running — α in fact grows with energy, and running it up on the one-loop estimate used throughout this paper, with the Standard Model’s charged fermions and no thresholds, puts its inverse near 50 at the Planck scale. Fuller treatments, which carry the thresholds and the W contribution pulling the other way, put it nearer 100.

Both are far from one, which is all this paragraph needs, and the same crude convention is used wherever this paper quotes a coefficient. What running supplies is the map. It does not supply the ultraviolet boundary value or the spectrum, and those are Road Thirteen’s price. D-EM.1 owes what all of this means for the corpus’s α, and the corpus claims no more than that the two point at one door.

What is left

One demand in five vocabularies: a second scale, dimensionless, fixed by S, B, R and C, that is not a count and does not contain α. The material face's second reading (Eleven); a maintenance cost in grains (Seven); a field-to-charge normalisation (Nine); a resource rule and a tolerance (Eight); the magnitude of the irreducibility S forces, as a fraction of cycle activity. These are one object in five dresses — AP24’s D1 throughout this paper, where the bare label is used — and Road Fourteen has put a sixth and better name on the same object: the response of the moving substrate to one grain. That name is worth more than the other five, because two items already open in the corpus may be the same thing said in their own vocabularies — AP15’s KS-31, the conserved balance between wave-side coherence and field-side connection that AP15 names as owed and does not exhibit, and AP43’s owed normalisation of its lean coefficient — not AP43-D1, which is the relativistic form and a different debt. Three open items in three papers, written apart and for different reasons, possibly one missing normalisation. The paper claims no more than that they are candidates for identification, and records them so that the next attempt starts there rather than at the counts. Nobody in this corpus has the object itself. AP24 §5.4 says it is not reachable from within, and says the check from without needs a normalisation, a spectrum or a boundary condition — which is Road Thirteen, with the spectrum as its price. This paper does not retract that and does not supply it.

One clarification about what is being asked for, because the demand above is easy to over-read. The missing object is the second scale, and a scale that already contains α would be circular. A selecting equation is a different thing: it will of course carry α as its unknown, and it may perfectly well carry the picture’s counts as coefficients. What it may not do is carry a coefficient whose size was set by knowing the answer. The test is the provenance of the coefficients, not their pedigree as counts.

What this paper adds to §5.4 is an accounting rather than a theorem. Twelve roads are closed, each for a stated reason. Two are open, each with its cost stated, and neither is this paper’s. And the class of forms that land is large enough that landing carries no information at all, which is §11’s business and the reason §11 exists.

A tolerance derived before any formula is that second scale. A tolerance set after a formula is the look-elsewhere effect wearing a band, and KS-30.4's 2,971 formulas become 2,971 confirmations. Derived first, or not at all.

What this section is not

It is not a proof that α cannot be derived. It is a register of what has been tried inside this corpus and why each attempt ended, together with the one structural reason most of them ended the same way. A route that contains something the corpus does not yet contain is untouched by all of it.

The corpus stands at one measured dimensionless input, as AP24 §5.5 left it. Which input it is may be chosen — Road Four shows the coupling and the gravitational ratio are interchangeable through AP28 — but the count of inputs is one either way, and no road in this register reduces it to none.