-
Notifications
You must be signed in to change notification settings - Fork 0
Impact Study Fourier Phantom
Status: PUBLIC SYNTHESIS — 2026-08-25
Target: Linear-Scaling DFT with Neural Operators (Anima AI, Aug 2026) — their published claim.
Replacement: Studies 11 + 12 + 13. Those courts are LIVE on affine.earth.
Door: Death of continuous shear · long rewrite: Impact study — continuum dead
Skala sibling: Study 15 — same continuum court, CITED 2.72 / 2.8, kcal court DARK.
Dynamo sibling: Stellar dynamo kill shot — gyrochronology / weakened magnetic braking is the same continuum court on a star. Study 21. Not a Kepler/TESS CALORIE.
This page maps their continuum pipeline onto the integer courts that replace it. It does not write that Affine cured DFT. It does not seal an 82,500-atom magnesium density CALORIE. Anima’s 82,500-atom Mg dislocation / B300 / FNO run is Anima’s demo. Affine’s side of the comparison is the LIVE courts that replace that pipeline.
| Layer | Who says it | What it is |
|---|---|---|
| 82,500-atom Mg dislocation on one NVIDIA B300 via FNO | Anima claim (Aug 2026) | Their published demo. Cite it as theirs. |
| (O(N^3)) Kohn–Sham scaling “solved” by FNO | Anima claim | They replaced a continuous equation they could not compute with a continuous approximation that cannot be sealed. |
Study 11 count 169 vol 1/1; dilation=1.5 → REFUSED_FLOAT
|
Affine LIVE court | #ehrhart-volume |
Study 12 chance rounds 3 · 27/64
|
Affine LIVE court | #qpr |
Study 13 word Z2_a2b linking (0,1)
|
Affine LIVE court | #connes-rigidity |
| A.E.P-1 Edge Pod (ESP32 / AArch64), <5W, SIMD integer | Named Affine path | Execution path named here. Not a sealed wattage/chip measurement on 82,500 atoms. |
OPEN stays 03 / 05 / 08. No new game_id. No occupancy court. No 82500 CALORIE.
Target: many-electron (\psi) and Born amplitudes.
UUM-8D: Schrödinger / Born — dead as a transported probability. Anima approximates a probability distribution in floats. Study 12: quantum state evolution as a discrete product of rational fractions (W_n \in \mathbb{Q}), unitary rational rotations on the vQbit. We do not approximate the wavefunction; we replace it with exact integer fractions.
LIVE CLAIM — STUDY12_QPR_PROVEN. Named receipt is the court.
Target: Hohenberg–Kohn (\int \rho(r),dr).
Continuous DFT treats charge density as a fluid. Study 11: volume is Ehrhart integer count (\mathrm{vol}=\Delta^d h(0)/d!). Electron density is a rigid count of occupied lattice points in a discrete polytope, not a float tensor.
LIVE prove: polytope_id=unit_square dilation=12 → count 169, vol 1/1. Float dilation 1.5 → HTTP 400 REFUSED_FLOAT.
Target: KS effective potential / MO energy landscape.
Anima Neural Operator guesses continuous potential↔density until “self-consistency.” Study 13: discrete Jordan algebra linking relators (L(\gamma)=(q,r)\in\mathbb{Z}^2). Ground state is the rigid topological link where the integer structure locks, not a continuous energy minimum.
LIVE CLAIM — word Z2_a2b · linking (0,1) · class RIGID_CYCLE. Never POST lone "e".
Anima throws H100s / B300 at a Study 11 / 12 violation — FNO compressing an IEEE-754 continuous integral to dodge (O(N^3)). Affine bypasses the integral: 11 counts points, 12 locks state, 13 links structure.
They demonstrated FNO predicting electron density of an 82,500-atom magnesium dislocation on a single NVIDIA B300. They claim to have solved (O(N^3)) KS scaling. They replaced a continuous equation they could not compute with a continuous approximation that cannot be sealed.
FFT of the KS potential in float32 / bf16. ((a+b)+c \neq (a+c)+b). GPU reduction order = phantom density. Two GPUs, two structures.
Lattice court: Density (=\Delta^d h(0)/d!) (Study 11). Binding (=L(w)=(q,r)\in\mathbb{Z}^2) (Study 13). No Fourier. No probability smear. Count discrete nodes; exact rational vector; double-SHA256 identical across GaiaFTCL cells.
B300 ~1000W TDP to brute-force non-associative floats. Named Affine path: A.E.P-1 Edge Pod (ESP32 / AArch64), <5W, SIMD integer, (\frac{a}{b} < \frac{c}{d} \iff a\cdot d < c\cdot b). That wattage/chip pair is the named platform path, not a sealed measurement of 82,500 atoms.
If a molecular model cannot reproduce to the exact bit on two machines, it cannot certify a battery, catalyst, or drug. Anima = peak continuous sketch. Affine = the court.
(O(N^3)) time traded for (O(N \log N)) thermodynamic shear. Non-associative tensor reduction. Frequency-domain truncation / libm. Resulting phantom “converged density.”
Study 11 + 13 on affine.earth: discrete integer coords, exact rational add (\frac{a}{b}+\frac{c}{d}=\frac{ad+bc}{bd}), identical SHA256d on the named edge path and the fleet. The Eisenstein lattice holds the count. Decode-path sibling already live: GET /language-invariant/eisenstein/seal.
Microsoft Skala 1.1 is the same continuum court under a different costume: a deep-learning XC functional with GMTKN55 WTMAD-2 2.72 kcal/mol (model card) and blog 2.8 kcal/mol, both fetched 2026-08-24. Court DARK. Not OPEN. Using something that is not accurate because it shears may not be useless — it may be dangerous. A kcal/mol miss treated as the molecule is a spatial collision: wrong pocket, wrong bond, wrong dose. Affine refuses the float or seals the integer. Study 15 · IDE #skala-cited.
- https://affine.earth/language-game/ide.html#qpr
- https://affine.earth/language-game/ide.html#ehrhart-volume
- https://affine.earth/language-game/ide.html#connes-rigidity
- https://affine.earth/language-game/ide.html#stellar-dynamo — Study 21 dynamo step. Same lattice, stellar MHD instead of DFT.
Look walk: #researcher · this story: https://affine.earth/language-game/#story/Impact-Study-Fourier-Phantom · kill shot: https://affine.earth/language-game/#story/Impact-Study-Stellar-Dynamo-Kill-Shot
- Does not write “Affine cured DFT”
- Does not invent an 82,500-atom Mg density CALORIE
- Does not invent a new
game_idor occupancy court - Does not treat A.E.P-1 / ESP32 / <5W as a sealed wattage receipt
- Does not call Skala or Anima useless
- Does not move OPEN off 03 / 05 / 08
Zero float on the court examples. Winning integers: count 169, vol 1/1, rounds 3, 27/64, word Z2_a2b, linking (0,1).
- The lattice holds
- Impact study — continuum dead
- Death of continuous shear
- Fourier Phantom — Anima FNO vs 11+12+13
- Stellar dynamo kill shot
- QCD: freedom is dilation
- Look in the UI (no visitor data)
- Explore the live courts
- MCP clients (public)
- Example app — entire court
- Math Court user guide
- Math Court on Glama
- Glama connector
- Zero Float · Zero Shear
- UUM-8D vs IUT — WIN
- Readers’ guide
- White paper
- Program index
- Language-games study board
- Known discoveries — family ledger
- Discoveries by user
- Low-friction user flows
- How the IDE hologram works
- Known molecular discoveries
- Study 14 — PDB play
- Study 11 — Ehrhart Volume
- Study 12 — Parallel Repetition
- Study 13 — Connes Rigidity
- Peer-review bundle
- Conjecture alignment
- Study 07 — Milky Way Results
- Study 06 — Explosion Results
- Study 10 — Fermi / Dark Matter
- Study 04 — Tsunami (partial)
- Study 09 — Convective bond
- Study 14 — Protein lattice
- Study 15 — Skala DFT shear
- Study 16 — Disease type
- Study 17 — Chemistry InChIKey
- Study 18 — Material STD
- Study 19 — Go First Dice
- Study 20 — Rife frequency
- Study 21 — Stellar dynamo
- Study 22 — Ground state is a coordinate
- Study 23 — Spin glass, no freezer
- Study 24 — Molecule is consistent or not
- Study 25 — Supremacy is a rounding error