A digital twin you play from where you actually are. No username, no password — your key stays in this browser.
We do not track people. Affine.Earth is a digital twin of the real world — your position is the coordinate you play from, the same way a piece needs a square on a board. It is bound once to your identity and is never a movement trail.
This creates a Bitcoin mainnet wallet in this browser at 0 BTC, credits genesis QFOT 100.0, and keeps the private key only here. You can download the key straight after.
100/1).Save your private key now. This is the only copy. Nobody can restore it for you — not us, not the mesh.
Every solver you have used declares optimality when an objective lands within a floating-point epsilon. You are asked to trust the tolerance. Affine returns an exact irreducible rational and a result that is IDENTICAL on nine independent machines — so the claim is checkable arithmetic, not a convergence promise.
Each answers at a URL on every cell. A 500 is the control arm — these are gates that can fail, which is the only thing that makes a pass mean anything.
reading the fleet…
The hardest thing to fake is an exact value past the machine word.
Sum 1/p over the first sixteen primes. The denominator is
32,589,158,477,190,044,730 — larger than Int64.max
(9,223,372,036,854,775,807), so no 64-bit rational can hold this state, and a
wrapping one returns a wrong number with no error. Check it in any arbitrary-precision
tool — Python: sum(Fraction(1,p) for p in primes) — and compare it
character-for-character with what the cell returns above. Identical, or we are lying.
Standard tensor-network truncation drops states below a float
threshold. That decision is a rounding artifact: it differs between libm versions, so two
machines replaying one contraction reach different states. Here viability is an integer
comparison — a·d < c·b on raw limbs — so the branch that dies is the same
branch on every cell, forever. The pruning decision is part of the proof rather than an
artifact of the arithmetic. The evidence is the number of branches sheared being
identical on all nine cells, above.
Growth is unbounded in REPRESENTATION LENGTH — a denominator may need hundreds of digits, and that costs bytes, not correctness. The surviving state count stays bounded by the invariant. Those are different axes, and conflating them is why people believe an exact substrate must eventually overflow. It does not; only its strings get longer. Memory is a single bump-allocated slab: allocation is a pointer increment, pruning is a pointer rewind, and fragmentation is impossible because nothing is ever freed piecemeal.
Scope, stated: these prove exact boundaryless arithmetic, deterministic discrete pruning, and allocation-free execution on the mesh. They are the substrate an NP-hard solver stands on. They are not a benchmark against a commercial solver on a named instance — that comparison is not run here, and this page will not imply it.
Enter to send · attach a schematic anytime · type flush to start fresh