Lattice OS
Engage
Noether · Lattice OS
Tw

Twin

The exact asset twin — any moment of the mission, in microseconds.
HOLD · mechanics validated vs NASA/JPL K=1012 in 49 µs

A digital twin is only as good as the physics inside it — and in cislunar space the physics is chaotic by construction. Halos, DROs, near-rectilinear halo orbits: their trajectories diverge from the first step, so a floating-point twin drifts, and a 15-year mission is millions of thermal cycles and billions of structural ones past where a simulation decorrelates. The industry's answer is a private model you re-run to trust and nobody else can audit.

Noether Twin holds the asset as an exact object. Its state at any cycle of the mission — past or future — is a lookup, not a run; its conservation ledgers hold as algebraic identities, not tolerances; and the whole thing ships as a certificate a third party re-derives from the file alone.

A conventional twin is a movie you play forward, drifting from frame one. This twin is an exact object — any frame, instantly, and provably the asset you certified.

THE TEST   Blind, against the cislunar industry's own reference — NASA/JPL's Three-Body Periodic Orbit Catalog (15,641 Earth–Moon orbits), with the open-source gold standard (DOP853 adaptive Runge–Kutta) as the parity bar.
THE RESULT   Rebuild the orbit, certify its stability, and get a warrant you can hand a partner, an auditor, or an insurer to check for themselves — without re-running anything:
  • Blind parity, at the gold standard. Ten seeded orbits (halo, Lyapunov, DRO), no tuning: worst stability-index deviation vs JPL 3×10−11, orbit closure 2×10−9, Jacobi drift 5×10−15.
  • The certificate the industry tools don't produce. A verified spectral bound on each orbit's monodromy contains the published JPL stability index, 110/110 across the family sweep, with a falsifiability bracket — and it is 521× cheaper to check than to re-derive.
  • The mission life, without the mission. The exact state at K=1012 cycles in 49 µs; the resonance envelope transported across 951 configurations at machine precision (2.6×10−16) where FEM time-integration of the same envelope is ~30,000 years.
  • Verifiable offline. The warrant re-derives from the file alone — no network, no apparatus — and a forged claim fails re-derivation. Certification is a state: defined events revoke it, and a safety limit cannot change quietly.
Log-log chart. The horizontal axis is mission cycles K from ten-cubed to ten-to-the-eighteenth. The vertical axis is wall-clock time. A grey dashed line, FEM time-integration at order K, rises steeply and passes a labeled dot at about thirty-two years at K equals ten-to-the-twelfth. A solid green line, the Noether Twin braid-clock fold at order log K, stays nearly flat near forty-nine microseconds across the whole axis. A vertical rule marks where float64 dies at step ten-to-the-fifth.
Certify the mission life without running it. Measured anchors (dots) on the scaling laws: FEM time-integration is O(K) — ~32 years at a decade-mission K=1012, and float64 loses the clock at step 105; the exact fold is O(log K), holding at microseconds across the whole axis. Different scaling laws, not a constant factor. FEM cost is a generous 1 ms/step lower bound.

The blind run — against JPL's own catalog

StepWhat was measuredThe number
The pinsthree orbit families fetched live and hashed — halos (the Gateway NRHO family), L1 Lyapunovs, DROs; μ taken from JPL's own system block15,641 orbits · sha-pinned
Blind parity10 seeded orbits rebuilt with the open-source integrator, no per-orbit tuning — closure, Jacobi drift, and stability index vs the catalogworst dev 3.2×10−11
The certificatea verified spectral bound per monodromy; the bound contains the JPL value and refuses just below it110/110 · both ways
Heat deaththe hottest swept orbit's growth exponent at K=1012, evaluated in O(1) where direct integration is ~1017 stages per orbitcertified · unreachable by stepping
The proofthe whole warrant re-derived from the file alone; a forged tighter bound fails on re-derivationoffline · one SHA-256

Twin steel thread & JPL case study — measured audits, blind-seeded, referee = the NASA/JPL Three-Body Periodic Orbit Catalog and DOP853. Demonstration of the capability across the catalog; the asset-specific twin is scoped per engagement.

How — the asset as one exact object

  1. Build from the spec, not the mesh. The asset's coupled periodic channels — orbit, thermal, structural modes, control loops — are held as one exact register with closed-form frequencies; no eigensolve, no time-marching. The engine is sealed; its mathematics is published. rebuilt to the recorded monodromy, byte-identical
  2. Any cycle is a lookup. The state at cycle K folds in O(log K) — the whole mission is random-access, exact, at microseconds — so the envelope across every configuration is transported by an exact scaling law, not re-integrated. K=1012 · 49 µs · transport dev 2.6×10−16
  3. The books balance as theorems. Energy and load-moment ledgers hold as exact algebraic identities under any re-steer — conservation that cannot drift, the audit the certificate rests on. conservation census exact · 300/300
  4. The answer is a certificate. Every claim re-derives from the file alone, offline; the certificate carries its own decertification triggers; issuance is mint-locked to a store only the substrate can produce — and no supercomputer or quantum machine can regenerate without it. verify · status · mint