Proposal — a new mechanism of gravitation:
entanglement drainage

This page proposes a cause. Not a field, not a geometry, not a bookkeeping metaphor — an earlier draft of this page leaned on one, and we retired it, because an accountant's description is not a mechanism. The cause: existing costs the vacuum entanglement, what is consumed must be drained, and the drainage is conserved. Gravity is the hydrology of monogamy. This is the theory the program was named after and — strangely — never tested directly until experiment forty.

emergent-gravity.com · July 2026 · builds on the main page and the impossibility triangle · the program's terminal synthesis: the drainage dichotomy · status: force sector measured — sign owned (n46) · clock sector refuted — headstone 20 (n47b) · flagship decoherence signature measured in the toy (n48) · not yet geometry, and says so

1 · The design rule: no axiom without a scar

This proposal was not free to be anything. Forty numbered experiments, seventeen headstones, two no-go theorems and one impossibility triangle each left a requirement that any surviving mechanism must satisfy. The inheritance table is the real foundation:

Measured wall	What it forces here
A medium has a rest frame — drag, momentum leak, no cone (n34)	the mediator is not a field in a substrate; it is a conserved flux pattern
A local scalar sector cannot be Newtonian and propagating (n35-A, 0/5000)	the long range must be protected by a conservation law, not by a pole
Nonlocality without gauge structure destroys causality (n36, measured)	the instantaneous piece must be organized by a constraint
A vector gauge flux holds the triangle but like charges repel (n37, measured)	the interaction must ride an even-rank response of the vacuum
Only energy-coupling is EP-compatible (n23–n28, measured)	consumption must track the dressing — the program's old m = dressing law
Records decohere, mediators must cohere — resolved by reading violations, not states (n39, measured)	the vacuum monitors sum-rule violations only

2 · The cause, in four steps

(i) Existing is expensive. A mass is not a thing placed in the vacuum; it is a pattern the vacuum maintains — a dressing. Dressing consumes local entanglement: the links a mass uses to exist are links the surrounding vacuum no longer has. This was this program's first measured law: m is the maintained dressing.

(ii) Monogamy is a conservation law. Entanglement is monogamous: each region has a budget. What a mass consumes cannot be created — it must be drained from elsewhere, through the network, link by link. A local sum rule holds at every site.

(iii) The Gauss law is the monogamy sum rule. A conserved drainage is a conserved flux: the flux of budget through any closed surface equals the consumption enclosed. This is not an axiom about gauge fields — it is what "nothing is consumed without being drained" means. The 1/r² of gravity is drainage dilution over spheres; the field is the catchment pattern; the potential is the depression of the water table.

(iv) Falling is catchment fusion. Two masses drain the vacuum. Apart, they deplete two full catchment basins; together, they share one. If the vacuum's response is an even-rank (scalar-like) susceptibility — and a critical vacuum's response to coupling patterns is exactly that — sharing is cheaper, and the gradient of the shared cost is an attractive force. Falling is two wells discovering they share the same water table.

3 · Experiment forty: the first gates, measured

The minimal implementation: a critical harmonic ring whose link couplings carry the budget; matter consumes a fraction of its sites' budgets; the drainage field is the minimal-cost rearrangement of couplings under the per-site sum rule (the cost law is the declared dynamical input); no force law, no Gauss equation, no sign is put in. measured script

What came out. The drainage profile of a single consumer is the compensated 1D Coulomb field — the jellium tent 0.1475·(1−2r/N), matched to three digits at every probed distance. The pair interaction is linear in distance with the ring's (½−d/L) curvature: the constant force of one-dimensional Newton, the same law the program's Gauss ladder measured in its old medium. The conservation law imposes an emergent global neutrality on the compact ring — a superselection rule arriving uninvited, exactly as a Gauss law on a torus demands. And the sign: like charges attract — because the interaction rides the critical vacuum's scalar susceptibility to the drainage pattern, an even-rank channel, where n37's vector flux could only repel.

That last line is the milestone. For the entire program, the two halves of gravity refused to coexist: mechanisms that attracted (the negative photograph, the relational bath) had no conservation law protecting their range; the mechanism with the conservation law and the causal cone (the minimal gauge model) repelled. Experiment forty is the first in forty where a conserved, Gauss-protected flux comes out attractive between like sources. One mechanism, both columns.

AMENDED — the sign claim did not survive n41 (headstone 19). When the dimension gate was run, re-running the committed n40 script did not reproduce the recorded slope: the vacuum-energy readout is repulsive, in 1D and in every dimension probed. The diagnostic that followed is sharper than the failure: the same configurations, read through the drainage-cost functional, are attractive in every dimension — and exactly Coulomb, to machine precision. The two energy functionals disagree everywhere, the cost term outweighs the vacuum term two-hundredfold, and the sign therefore belongs to whichever functional is declared to be the physical energy. Attraction was an input hiding in the declared cost law, not a discovery. The milestone is demoted to what was actually measured: a conserved, Gauss-protected, emergently neutral Coulomb response — in one, two and three dimensions. The mechanism has gravity's form; it borrows gravity's sign, and will own it only when the cost law is derived from the vacuum's own susceptibility instead of declared. That derivation is now the program's single most important open problem. pre-registration + results · script

UPDATE — the sign is owned (n45, n46). The derivation was attempted both ways, gates written first. The energy route died by a theorem: the vacuum's zero-point energy is concave in its couplings — it cannot price the drainage, in any dimension, ever (n45). The information route passed every gate it faced: take the cost to be the vacuum's own information metric — the fidelity metric of the Gaussian ground state, positive by construction, nothing declared — and the interaction between like consumers comes out attractive in one, two and three dimensions: machine-exact ring Coulomb in 1D, an exact staggered zero mode everywhere, quadratic infrared dispersion in 2D and 3D (0.24% and 0.36%). The dressed stiffness of that dispersion is the toy's 1/G: 202 in 2D, 493 in 3D, bare-lattice units — the first number of the constants program. Two registered gates fired along the way and are recorded with everything else: the 2D bare-Green form gate (a UV-contaminated reference) and the 3D dispersion ratio at its registered lattice size (5.8% vs 5%, converging as 1/L² to 0.36% — the lattice was miscalibrated, and an honest replication at pre-registered L=96 is owed as n46c). Said plainly: gravity's sign is no longer borrowed. Existence costs the vacuum entanglement; monogamy conserves the budget; the price of rearrangement is the distance the vacuum's state must move; and two masses are cheaper together than apart — in information, not in energy. The one sentence survives with its verbs measured. n45 · n46 · n46b (infrared) · script 1D · 2D · Fourier

4 · The quantum layer

What does the vacuum read? In this proposal: only violations of its own sum rule — never the states. The exact quantum toy of this structure (a gauge theory whose stabilizers are monitored) showed that such reading preserves a dressed mass superposition to 1.00000000 while any field-reading vacuum destroys it at the predicted rate. measured script The drainage vacuum is therefore a quantum channel, and the proposal inherits the cleanest fork this program owns:

BMV comes out positive — gravitationally mediated entanglement survives, in direct opposition to this program's own book two (the field-reading vacuum: BMV-null, Diósi–Penrose floor). Both vacua agree on Newton; the 2030s experiments read which one we live in. Kill: a confirmed BMV null.
No decoherence floor at rest; an inverted Diósi–Penrose at high frequency — decoherence only when matter outruns the drainage. measured in the toy (n48): the static residual decays through eighteen orders of magnitude — no floor, where DP demands one; the rate switches on as the exact Lorentzian (dT₀₀/dt)²/(γ² + ω²)·γ⁰… with knee at the drainage rate; and the low-frequency law is rate ∝ (dT₀₀/dt)²/γ_drain² — the exponent measured, correcting this page's earlier /γ. A frequency-swept levitated-mass interferometer sees: nothing (GR), a floor (DP), or this knee (drainage). Kill: a static DP floor. pre-registration + results · script

5 · Predictions and executioners

BMV positive (above). the fork
Inverted-DP high-frequency decoherence (above). proposed
3D drainage dilution gives 1/r² — the flux-conservation argument is dimension-blind; the 2D and 3D lattices must deliver 1/r and 1/r² with the same attractive sign. The gate was run (n41, pre-registered): the form half passed spectacularly — lattice Coulomb in 2D at 0.44% against a 0.7% gate, exact Coulomb in the cost functional in every dimension — and the sign half fired the death clause on the registered observable. See the amendment in section 3: the sign was the declared cost law's, not the vacuum's. form: measured, every D sign: headstone 19
Drainage has a speed — the rearrangement propagates at the vacuum's entanglement-propagation velocity: one substrate, one cone, one c for field and matter. The dynamical version of n40 must show the cone; its absence kills the relativistic reading. argued
Galaxy lopsidedness aligns with absolute motion — refuted, headstone 18. The drainage wake predicted that a galaxy's fast rotation side aligns with its peculiar-velocity direction (sign and radial profile fixed by the wake toy before the test). We then ran the measurement ourselves, pre-registered before any data was opened: 236 WALLABY HI velocity maps (CADC), side-asymmetry D extracted with an estimator validated on displaced-well mocks, full 3D peculiar velocities from the 2M++ reconstruction. Result: N = 111 galaxies passing all pre-registered cuts, fast-side/leading-side agreement 50.5% — a coin flip (binomial p = 0.50; amplitude correlation ρ = −0.02). The null gate (p > 0.1 at N ≥ 100) fires as written: the prediction is dead at pilot level, and it takes the drainage-lag reading of merging-cluster lensing offsets with it, as pre-registered — same mechanism, same grave. AMENDED after a reader's challenge forced the power analysis the pre-registration should have contained: with the measured intrinsic scatter (σ_D = 0.31, a cluster-contaminated sample), N = 111 only excludes agreement ≥ 60% — i.e. only the STRONG-wake regime (D_wake ≳ 0.15). The realistic regime (D_wake ~ 0.02–0.05) needed N ~ 1,100–11,000 and remains untested: the full WALLABY survey or MaNGA can decide it. So the epitaph, corrected: dead is the strong coupling and the un-powered pre-registration; the moderate wake — and the Bullet-lag reading at its natural amplitude — are wounded, not buried. New program rule, learned here: no pre-registration without a minimum-detectable-effect calculation. The static sector (n40) was never touched either way. pre-registration · pipeline · the killed wake model
Λ is the idle drainage rate — the cost of maintaining the vacuum's own correlations, the residual metabolism with nothing to feed. quantitative form open

6 · The proposal's own honest ledger

Measured (each in an exact toy): the m = dressing law (the program's first result); conservation of the monitored charge under relational reading (n31); force, propagator and back-reaction from one bath (n33); the gauge cancellation outside the cone to 69 decades (n37); code-space protection under violation-reading (n39); and now the drainage tent, the neutrality, the conserved Coulomb response in one, two and three dimensions (n40, n41), and — the program's deepest result — the attractive sign from the vacuum's own information metric, in every dimension, with nothing declared (n46). Derived, no longer declared: the cost law. The energy route died by a theorem (n45: zero-point energy is concave in the couplings — it can never price the drainage); the information route is the theory's founding sentence taken literally, and it passed. To be constructed:

~~n46c: the 3D infrared replication owed~~ — paid: pre-registered L=96 gates both pass (0.36% vs 5%; convergence 0.36% vs 2%), S₃D = 493.44. registration + results;
local clocks — the metric reading is dead: headstone 20 (n47b). Clocks do slow near a mass, universally in sign — but the response is quadratic in the source, and the pre-registered hunt for a linear channel found none: the apparent one was higher-order leakage, exposed by the control clock; the two spectral clocks' linear coefficients sit at the systematic floor with opposite signs. The failure is structural, not numerical: the drainage field is purely staggered, and a purely staggered field offers parity-blind clocks only its square. GR's redshift is linear in the source and is measured — GPS falsifies the static mechanism's clock sector as it stands. Resurrection requires a smooth (q≈0) drainage component coupled linearly to consumption; nothing in the current model provides one. This is now the mechanism's deepest wound, stated at full prominence. Second-order update (n50): the state-dependence of the information metric does open a smooth channel — local binding strengthens measurably in the well — but it responds as M², not M: the parity obstruction survives every order, because cell-averaged observables are even in the staggered field by symmetry. Conjectured theorem (n51a): on a bipartite substrate no parity-neutral observable can respond linearly to mass. The architectural escape was run the same day — and the fork closed (n51): the drainage dichotomy is measured from both sides. The signed architecture (Gauss sector at q=0) delivers exactly linear clocks — F-exponent 1.000, every construction — but like masses repel (machine-Coulomb), and even the linear clock response reads the local flux, not the potential. The root cause is one sentence: in general relativity the potential is itself a local field (g₀₀); in any flux-conservation drainage the potential is a nonlocal integral of the flux, and local matter can only couple to the flux. Potential-coupled clocks — gravitational redshift — cannot arise in this class of theories, in either architecture. Geometry is closed for quadratic drainage. What survives, and it is not nothing: Newton's force with an owned sign, an emergent conservation law, a first constant, and a laboratory signature (the decoherence knee) that no metric theory shares. What died: the hope that this same bookkeeping would hand over Einstein's clocks. The dichotomy is now a theorem (n51a): the unsigned architecture's only long-range channel sits at the staggered point, where every parity-neutral observable carries an exact filter zero — in one dimension the linear response vanishes identically, at machine precision; and in the signed architecture, Bochner's theorem forces repulsion for every positive cost, exactly (E_int = f²K(d), verified to 10⁻¹²). The static program's structural map is complete. the proofs + verification · gates n50 · n51 + results · script n47 · n47b + results · script;
the parity exposure, made precise (n49): same-parity consumers anti-gravitate — machine-exact repulsive Coulomb, measured. The model therefore requires that all ordinary matter be one parity class, the other class belonging to the vacuum's compensation sector — now the mechanism's load-bearing interpretive assumption, named. And the short-range signature table: parity-oscillating deviations from Newton, decaying ~1/r, exact numbers published. pre-registration + results · script;
dynamics: the drainage cone, retardation, and the triangle held without a posited constraint;
EP: consumption ∝ dressing ∝ energy, closing the loop with n23–n28's requirement — universality as a theorem of the budget;
~~the staggered carrier: a possible artifact-source~~ — resolved by n41's control: in 2D and 3D the carrier is no longer dominantly staggered (43%, 7.5%) yet the Coulomb form and emergent neutrality hold unchanged; the alternating mode was the 1D representation of the flux, not its cause;
numbers: G from the drainage stiffness, in physical units.

Nearest neighbors, named: induced/entropic gravity (Sakharov, Jacobson, Verlinde), entanglement-thermodynamics (Van Raamsdonk, Lashkari–Van Raamsdonk, Faulkner et al.), holography-as-error-correction (Almheiri–Dong–Harlow). What is distinct here and must survive referees: the conservation law doing the work is monogamy itself, imposed as a local sum rule on couplings rather than derived from horizon thermodynamics; the mechanism is a drainage flow with its own declared dynamics, sign and force law measured rather than postulated; and it ships with a falsifiable fork against a fully-developed rival from the same research program. If prior art for this assembly exists, this page will say so, per the rules of this site.

7 · One sentence

Weight is what your existence costs the vacuum: every mass is a standing withdrawal from a shared account of entanglement, the Gauss law says no withdrawal goes undrained, and falling is two wells discovering they share the same water table. The rest of this site is the forty-experiment record of how we were forced to exactly this sentence — and the experiments of the next decade get the signature.