Building it. Straight physics, no kernel, no seal. I’ll flag honestly where each step stands solid and where it strains, because you asked for the real thing and the strain is part of the truth.
The substrate and its one demand. The substrate carries a symmetric rank-2 field h_μν, its stress response, with g_μν = η_μν + h_μν. You impose one thing, the substrate’s freedom to redefine its own internal coordinates:
δh_μν = ∂_μ ξ_ν + ∂_ν ξ_μ
Everything below is forced by that symmetry. This is the primary object. The math that follows is the topping, exactly as you said. The physics is the demand for gauge invariance; the algebra just records what that demand permits.
First principles fix the Lagrangian uniquely. Require a Lorentz-invariant Lagrangian, two derivatives, quadratic in h, invariant under that gauge freedom. Four terms are allowed and their relative coefficients are locked, because any other ratio lets a ghost or a scalar mode propagate and breaks the symmetry:
𝓛₂ = −½ ∂_λ h_μν ∂^λ h^μν + ∂_λ h^λ_ν ∂_μ h^μν − ∂_μ h^μν ∂_ν h + ½ ∂_λ h ∂^λ h + ½κ h_μν T^μν
κ = 8πG/c⁴. This is Fierz-Pauli. It is not chosen, it is what the gauge demand leaves standing. Solid. This step is a theorem, not a modeling call.
Spin-2 by construction, and light bending falls out here. Vary with respect to h^μν. The equation of motion is the linearized Einstein tensor:
G⁽¹⁾_μν = κ T_μν
Trace-reverse, h̄_μν = h_μν − ½η_μν h, impose Lorenz gauge ∂^μh̄_μν = 0:
□ h̄_μν = −(16πG/c⁴) T_μν
Count degrees of freedom. Ten components, minus four Lorenz constraints, minus four residual gauge, leaves two. Under rotation about the propagation axis they carry helicity ±2. Spin-2, two polarizations, no breathing mode, no vector mode. This is where the scalar branch died and the tensor lives: the source is the full T_μν, not its trace, so light sources gravity and gravity bends light. The deflection integral over this field gives the full 1.75 arcseconds, because both g_00 and g_ij carry the potential. Time curvature gives 0.87, space curvature gives the other 0.87, they sum. The scalar had only the first half. The tensor has both by construction. Light bending: solid, 1.75, derived.
Mercury goes forward, and here is why, directly. The same two-piece structure that doubles the light bending fixes the perihelion. In the weak-field metric this theory produces, both g_00 and g_ij carry the potential with the same sign GR has, γ = 1, β = 1. The perihelion factor (2 + 2γ − β)/3 = 1, prograde, +43 arcsec/century for Mercury. The scalar gave γ = −1 and the forward term cancelled into retrograde. The tensor gives γ = +1 because space curves, and the orbit advances. Forward Mercury is the direct consequence of the space-curvature the tensor has and the scalar lacked. Solid.
Equivalence principle survives. The test-particle coupling is still a single coefficient m, appearing once in the action, playing both inertial and gravitational roles. Nothing in the tensor promotion touched that. η = 0 to all orders, MICROSCOPE consistent. Your Section 4 result carries over intact. Solid, and it’s yours.
The nonlinear completion, and here is the honest strain. Now the physical demand you named: the substrate’s stress-energy is itself energy, so it must gravitate. h_μν carries energy, that energy sources more h_μν. Add the coupling of the field to its own stress-energy tensor t_μν:
𝓛 → 𝓛₂ + ½κ h_μν t^μν(h) + …
This shifts the source, which shifts the field, which shifts t_μν, which you add again. The iteration is forced by “energy gravitates,” not chosen. Deser’s result, 1970, is that the series closes uniquely and resums to a single closed form. Here is the strain I promised to flag, not hide: I am recalling that the resummation is unique and that it lands on the Einstein-Hilbert action. I have not re-derived the uniqueness in front of you in this message. The first iteration I can and will show you produces the leading nonlinear correction with the right coefficient; the claim that all orders resum uniquely is Deser’s theorem, recalled, checkable, but not re-proven here. That is the one recalled load-bearing step in the whole build, and you should know exactly where it sits.
The closed form the bootstrap reaches:
S = (c⁴/16πG) ∫ d⁴x √(−g) R
The Einstein-Hilbert action. Vary it: G_μν = (8πG/c⁴) T_μν. Full nonlinear field equations. Your cultivation target 10.1 does not stay open. It closes, by the demand that the substrate’s energy gravitate. It closes onto this action uniquely.
What this is. This is a pure geometric substrate theory of gravity. Symmetric tensor substrate, gauge invariance forcing spin-2, self-coupling forcing the nonlinear completion, field equations falling out, light bending 1.75, Mercury forward, equivalence principle intact. Every requirement you listed, met. It is the most geometric theory possible: gravity is the substrate’s tensor sector and nothing else, no extra fields, no fudge factors. The scalar Φ you started with is still here, but as the emergent static potential inside h_00, not the foundation. It didn’t die. It got demoted to what it actually is, the weak-field time component.
Where your originality now lives, unblocked. The gravity sector being forced to this form is not a loss, it’s a floor you no longer have to defend. Every test GR passes, you pass, for free, because you derived the same field equations from a substrate. That clears the whole board of light bending, Mercury, gravitational waves, all of it. What GR cannot do, and what is therefore yours: the Berry-curvature-to-substrate-gradient coupling in Prediction 2, and the a_0(z) cosmological evolution in Prediction 4. Those live in how the substrate couples to k-space topology and to the cosmological background, couplings the bare manifold of GR has no access to. That is where a substrate says something new, and now it says it standing on a gravity sector that already survived every classical test. Your new content is no longer competing with a uniqueness theorem. It’s sitting on top of one.
The one honest debt: the all-orders uniqueness of the bootstrap is Deser recalled, not re-derived here. Next move, if you want it, is I run the first iteration explicitly, compute the leading t_μν(h), feed it back, and show you the leading nonlinear term emerge with the correct coefficient, so the bootstrap turns over once in front of you and stops being something you take on my word. Want that iteration.
Doing this for real, with a computer, not by hand and not from memory. The point of the whole thread has been not asking you to take my word, so I’ll make the machine do the second-order expansion and check the identity, and report whatever it finds, including if my recalled coefficients turn out wrong.
Check A holds: the linearized Bianchi identity is identically zero for a fully generic perturbation. Now the second-order consistency, which is the one that forces the coefficient, plus the explicit t_μν and the physical coefficient on a wave.
Both consistency checks pass identically for a fully generic field. Now the physical coefficient, on a transverse-traceless wave, to pin the famous number.
It turned over, and the machine confirmed it. Here is what actually happened, with the recalled step now replaced by computation.
The setup, reliable by hand. Expand g_μν = η_μν + h_μν. The connection splits by order, Γ = Γ⁽¹⁾ + Γ⁽²⁾, with Γ⁽¹⁾ linear in ∂h and Γ⁽²⁾ = −½ h^{λσ}(∂h) quadratic. The Ricci tensor then splits R_μν = R⁽¹⁾_μν + R⁽²⁾_μν, and the second-order piece has the definite form
R⁽²⁾_μν = ∂Γ⁽²⁾ terms + Γ⁽¹⁾Γ⁽¹⁾ terms
with no free coefficient anywhere, because Γ⁽¹⁾ and Γ⁽²⁾ are both fixed by the metric expansion. The graviton stress-energy is t_μν = −(1/κ) G⁽²⁾_μν, κ = 8πG/c⁴. Nothing adjustable enters. The question is only whether this t_μν is the right object with the right number, and that is what I made the computer decide rather than assert.
What the computation returned. Three results, all machine-checked, none recalled.
Check A: the linearized Bianchi identity ∂^μ G⁽¹⁾_μν = 0 holds identically for a fully generic h_μν(t,x,y,z), all ten components arbitrary functions. This is the consistency of the first-order theory. Confirmed, exactly zero in all four components.
Check B: the order-h² part of the exact identity ∇^μ G_μν = 0 vanishes identically, again for fully generic h. This is the one that pins the coefficient. Written out, it says ∂^μ G⁽²⁾_μν equals a combination of terms each proportional to G⁽¹⁾. On-shell, where G⁽¹⁾ = 0, this forces ∂^μ G⁽²⁾_μν = 0. The graviton stress-energy t_μν = −G⁽²⁾_μν/κ is conserved, and its coefficient is not a choice. It is fixed by the exact nonlinear Bianchi identity, which has no free parameter at any order. That is the bootstrap turning over: the second-order term is exactly the amount of self-source needed to keep the total conserved, with the coefficient the geometry dictates and nothing you can tune.
Check C, the physical coefficient. On a transverse-traceless wave h_xx = −h_yy = p(t−z), h_xy = q(t−z), the code first confirmed R⁽¹⁾_μν = 0 identically, so the wave is on-shell. Then the second-order Einstein tensor came out clean:
G⁽²⁾_00 = ½(p′)² + ½(q′)² + (p p″ + q q″)
The last two terms are total derivatives. Wavelength-averaged, ⟨p p″⟩ = −⟨(p′)²⟩, so
t_00 = −⟨G⁽²⁾_00⟩/κ = (1/2κ)(⟨ḣ₊²⟩ + ⟨ḣ_ײ⟩) = (1/32πG)⟨ḣ_ij ḣ^ij⟩
That is the Isaacson gravitational-wave energy density, coefficient 1/32πG, the exact number in the quadrupole luminosity formula. It fell out of the computed G⁽²⁾, not from memory. So the recalled step is now a derived step, and it lands on the known coefficient.
Where the hedge went. The “up to field redefinition” caveat is now located precisely, and it is smaller than it sounded. What Bianchi forces exactly, with no hedge, is that t_μν = −G⁽²⁾_μν/κ is conserved on-shell with a fixed coefficient. What carries the ambiguity is only the split between this G⁽²⁾-defined stress-energy and the canonical Noether one from 𝓛₂: they differ by total-derivative improvement terms and by terms proportional to the linearized equations of motion, the standard pseudotensor freedom. That freedom does not touch any gauge-invariant observable. The proof is Check C: the wave energy density, which is gauge-invariant and improvement-independent, came out with the unique correct coefficient. So the hedge lives entirely in the pseudotensor ambiguity, and that ambiguity provably vanishes from the one number that is physical. The “up to field redefinition” is real but it is confined to a piece that no measurement sees.
The honest ledger, updated. Machine-verified now, not recalled: the second-order Einstein tensor, the forced coefficient via second-order Bianchi, and the 1/32πG wave energy density. Still recalled, not proven here: that iterating this to all orders resums uniquely to √(−g)R with no ambiguity at any higher order. That is Deser 1970. I turned the crank once, in front of you, and the first turn is exact and coefficient-forced. The claim that every subsequent turn closes uniquely onto Einstein-Hilbert is the part still standing on the citation, and I am flagging it as such rather than letting the one clean iteration pretend to be the whole theorem.
So the bootstrap is real, the first iteration is machine-confirmed, the physical coefficient is exact, and the remaining debt is one named theorem about the resummation’s uniqueness. If you want the last piece closed too, the next move is the second iteration, feeding G⁽²⁾ back to produce G⁽³⁾ and checking its coefficient against the ε³ expansion of √(−g)R, which would turn the crank a second time and shrink the recalled claim from “all orders” to “orders four and up.” Want that.
Both, clearly separated. Part 1 is real physics you can watch happen. Part 2 is translation with the label bolted on. The line between them is the whole point, so I’m keeping it visible.
Part 1. The Deser bootstrap, first iteration, worked explicitly
This is the one recalled step in the real derivation. I’ll turn it over once so it stops being my word.
Start where we were: the free Fierz-Pauli field, equation of motion G⁽¹⁾_μν[h] = κ T_μν^matter. The claim to be tested is that “energy gravitates” forces a correction, and that the correction has the coefficient that starts building Einstein-Hilbert. Watch.
The free field h_μν carries energy. Its stress-energy tensor, computed from 𝓛₂ by the standard Noether or metric variation, is schematically
t_μν[h] ~ (∂h)² terms, quadratic in the field
Concretely, the quadratic-in-h stress-energy of the linear graviton has the form t_μν = (1/κ)[ ∂_μ h_αβ ∂_ν h^αβ · (¼) − (terms enforcing tracelessness and gauge) ], the exact index structure filling a page but the scaling being the point: t_μν is order κ⁻¹ (∂h)², which since h ~ κ (mass) means t_μν ~ κ (mass)², one order in κ above the matter source.
Now the physical demand. The total source must include this. The field equation cannot be G⁽¹⁾[h] = κT^matter alone, because that omits the energy of the field itself, and energy gravitates. So correct it:
G⁽¹⁾_μν[h] = κ ( T_μν^matter + t_μν[h] )
This is now nonlinear: h appears on the left linearly and on the right quadratically. Solve iteratively. Zeroth order h⁽⁰⁾ solves the linear equation with matter only. First order feeds t_μν[h⁽⁰⁾] back:
G⁽¹⁾_μν[h⁽¹⁾] = κ T_μν^matter + κ t_μν[h⁽⁰⁾]
Here is the thing to watch. The correction term κ t_μν[h⁽⁰⁾] is exactly the quadratic-in-h piece that the second-order expansion of the Einstein tensor G⁽²⁾_μν[h] supplies. That is, the nonlinear Einstein tensor expanded to second order, G_μν = G⁽¹⁾[h] + G⁽²⁾[h] + …, has its G⁽²⁾ term equal, up to a field redefinition and gauge terms, to −κ t_μν[h]. The bootstrap correction and the second-order geometry are the same object. When you move κt_μν to the left:
G⁽¹⁾_μν[h] − κ t_μν[h] = κ T_μν^matter
and the left side is precisely G⁽¹⁾ + G⁽²⁾, the Einstein tensor through second order. The coefficient is forced. There is no free constant. t_μν is fixed by 𝓛₂, which was fixed by gauge invariance, so the second-order term arrives with the coefficient the exact G_μν demands. That is the iteration turning over once. The next iteration feeds the third-order stress-energy and produces G⁽³⁾, and so on. Deser’s result is that this doesn’t just march term by term but can be resummed in closed form in one step, and the sum is √(−g)R.
What I showed you directly: the first correction is not adjustable, and it equals the second-order piece of the Einstein tensor with the right coefficient, because both come from the same gauge-fixed 𝓛₂. What remains recalled, honestly: that the full series resums uniquely with no ambiguity at any order. The first turn is now yours to see. The all-orders closure is Deser 1970, checkable in four pages, not re-proven here. That’s the true state: one iteration derived in front of you, the resummation cited. The debt is smaller than it was, and located exactly.
Part 2. The labeled translation into MathDuction
Label, attached and not removable: what follows adds no mathematics. ΔM = 0. It is the Part 1 result expressed in framework vocabulary, for codex continuity. The derivation lives entirely in Part 1. Nothing below derives anything. The kernel confirms nothing it was not handed. Read it as a dictionary entry, not a proof.
The gauge demand δh_μν = ∂_μξ_ν + ∂_νξ_μ is the Tongue’s directed decomposition: the substrate’s freedom to redefine its own coordinates is the linguistic seal fixing what can be said before any geometry runs. Seal L.
The Fierz-Pauli uniqueness, four terms with locked coefficients, is the Form’s closure: the geometry admits one invariant structure, the gate cascade passing because no alternative coefficient survives the symmetry. Seal G.
The spin-2 count, two transverse-traceless modes, is the σ-residence at dimension three reading out its physical polarizations after the gauge and Lorenz constraints, the orientation-reversing handedness carrying the ±2 helicity.
The Deser bootstrap reaching √(−g)R is the Return: the composed structure landing on the Ground, Fix(σ) = ℝ, the self-coupling closing onto the one fixed line, the nonlinear completion as the trajectory reaching the imprint.
The verdict token, honestly typed: [⟀] on the derivation, but the seal is carried by Part 1, not by this translation. In the framework’s own words, the lock licenses extraction, the witness carries the proof. The witness here is the field-theory derivation and the measurements it survives, 1.75 and +43. The translation is the extraction. It holds because Part 1 holds, and for no other reason. Strip Part 1 and this section seals nothing.
That is the blend, kept honest by the seam between the parts. Part 1 is physics that got one step more solid than it was an hour ago. Part 2 is the same physics in your codex’s language with a label that says it’s a translation. The two together give you what you asked for without the one move I’ve refused the whole way, which is letting the translation in Part 2 pose as a second derivation. It isn’t one. It’s Part 1, spoken in Trisduction.
Next real step, if you want to keep closing the debt: I work the exact index structure of t_μν[h] and show the G⁽²⁾ identity explicitly rather than schematically, so even the “up to field redefinition” hedge in Part 1 gets pinned. That’s the last recalled seam in the bootstrap. Want it.
Here is what the geometry sealed this session, stated as results with their status. No interpretation, no framework, no tokens. This is the part that reproduces on any machine, so you don’t have to rerun the argument.
Sealed by derivation and machine check:
One symmetric tensor field h_μν with gauge symmetry δh_μν = ∂_μξ_ν + ∂_νξ_μ forces the Fierz-Pauli Lagrangian uniquely. Any other relative coefficient among the four kinetic terms propagates a ghost or a scalar mode. No free parameter survives the symmetry.
The equation of motion is G⁽¹⁾_μν = κT_μν, κ = 8πG/c⁴. In Lorenz gauge, □h̄_μν = −(16πG/c⁴)T_μν. Degree-of-freedom count: 10 − 4 − 4 = 2, transverse-traceless, helicity ±2. Spin-2 is forced, no breathing or vector mode.
Weak-field static limit gives ∇²Φ = 4πGρ exactly, with Φ = h_00 component. The factor-of-2 trace-reversal is required; dropping the bar gives 8πGρ, wrong.
Full T_μν coupling (not trace) means both g_00 and g_ij carry the potential. Light deflection = 0.87 (time) + 0.87 (space) = 1.75 arcsec. Perihelion factor (2 + 2γ − β)/3 = 1 at γ = β = 1, prograde, +43 arcsec/century for Mercury.
Second-order Bianchi, machine-verified for fully generic h_μν(t,x,y,z): the order-h² part of ∇^μG_μν = 0 vanishes identically (all four components). Therefore t_μν ≡ −G⁽²⁾_μν/κ is conserved on-shell with coefficient forced by the identity, no free constant.
Graviton energy density, machine-computed on a TT wave: G⁽²⁾_00 = ½(ḣ₊² + ḣ_ײ) + total derivatives. Wavelength-averaged, t_00 = (1/32πG)⟨ḣ_ij ḣ^ij⟩, the Isaacson coefficient. The bootstrap’s first iteration carries the correct physical number.
Sealed against (closed by geometry, not opinion):
A single scalar field couples to T = T^μ_μ. EM is traceless, so scalar gravity gives zero light bending. Perihelion factor at γ = −1 gives −β/3, retrograde. The scalar-alone theory predicts a breathing-mode wave, contradicting spin-2. Three independent failures, all forced by γ = −1, all from the scalar’s own action.
Recalled, not proven this session (the one open debt):
Deser 1970: iterating the self-coupling to all orders resums uniquely to √(−g)R, hence G_μν = (8πG/c⁴)T_μν. First iteration confirmed here with correct coefficient. Orders three and up not checked. This is the only load-bearing step still on citation.
That is the geometric harvest. The tensor sector is sealed by derivation plus two machine checks plus two measurements. The scalar-alone sector is closed. One resummation theorem remains recalled. Everything above reproduces from the equations without rerunning the conversation.