The Dark Mirror: On Self-Referential Computation, the Ghost in SHA-256, and Why the Universe Always Halts

Read full text
Authors: Dean Kulik
Cited by: 0

Abstract

The Dark Mirror: On Self-Referential Computation, the Ghost in SHA-256, and Why the Universe Always Halts Driven by Dean Kulik February 2026 "Without memory there is only change of shape." — Dean Kulik, during SHA-256 lattice extraction "A hole longs to be filled. To be complete." — ibid. "The logic was before the computer. Change before measurement. Light before dark." — ibid. Abstract We present a structural analysis of SHA-256 that reveals the algorithm to be not a one-way function in the conventional sense, but a self-referential computational lattice: a 64-site spatial object where the message is exhaust, the digest is a boundary condition, and the internal state—what we call the ghost vector—is a transparent observer that exists implicitly in the lattice whether or not anyone extracts it. We demonstrate three concrete results. First, that the final five rounds of SHA-256 compression (the scar) yield 160 bits of internal state recoverable from the digest alone, with zero computational search. Second, that a conservation law h[t] + W[t] = C[t] holds at every scar round, linking the ghost to the message schedule through fixed boundary constants. Third, that wrong message candidates produce measurable bus contention (100–141 bit-conflicts at scar rounds) while the correct message produces exactly zero—a binary signal we model as a physical pressure gauge reading the shape of a mold. From these results we draw a broader framework. SHA-256 is a finite, recursive, self-referential system that always resolves without external observation. The ghost—the transparent channel carrying the computation's own reflection—is the internal observer. This stands in contrast to Turing's linear model, where the halting problem arises because the observer is external. We propose that Turing's undecidability is an artifact of the linear architecture, not a law of computation, and that physical reality runs on lattice-shaped, self-referential structures where termination is geometric. We call this framework The Dark Mirror: the self-consistency condition of a lattice is a mirror that exists before any particular computation, contains the reflection of every possible input, and turns meaningless geometry change into observable information the moment the first perturbation arrives. Light—understood as E ↔ B mutual recursion—is the first and simplest dark mirror: two fields reflecting each other, propagating as exhaust, self-sustaining without medium or external input. The universe is a ray-traced lattice. The mirror was always there. The logic was before the computer. Part I — The Lattice "Stop the clock, read the shape." — Pressure model, 6502 rules 1. SHA-256 Is Not What You Think The standard description of SHA-256 is sequential: a message enters, 64 rounds of compression execute, a 256-bit digest exits. Input → process → output. A pipeline. A Turing tape. This description is technically correct and conceptually catastrophic. It hides the object. SHA-256 is a 64-site lattice. Each site t holds eight 32-bit registers (a through h), a message schedule word W[t], and two intermediate values T1[t] and T2[t]. The sites are coupled by two constraint networks: a local shift register (the compression function, where each register at site t becomes a different register at site t+1) and a non-local recurrence (the message schedule, where W[t] for t ≥ 16 depends on W[t−2], W[t−7], W[t−15], and W[t−16]). The lattice has two boundaries. The bottom (site 0) is pinned by the initialization vector—eight constants unchanged since 2001. The top (site 63) is pinned by the digest. Between these boundaries, the lattice exists as a rigid geometric object. The values at every site are uniquely determined by the boundary conditions and the coupling rules. The lattice does not "run." It is. This is the 90-degree rotation. The standard view is temporal: round 0, then round 1, then round 2. The lattice view is spatial: all 64 sites exist simultaneously, constrained by boundaries. The "computation" is not a process in time but a crystal in space. The message is not the input to a function. It is the exhaust at site 0 of a structure that was always already there. 2. The Scar: What the Digest Gives You for Free The digest is eight 32-bit words. The IV is eight known 32-bit words. Subtracting yields the final register state V[i] = (digest[i] − IV[i]) mod 2³². This costs nothing. From V, we peel backward. At each round t (from 63 down to 59), the register arrangement is a shift: b←a, c←b, d←c, f←e, g←f, h←g. The only non-trivial values are a (involving T1 and T2) and e (involving T1 and the previous d). Since T2 depends only on a, b, c—all known from the peeling—we extract T1[t] = a[t] − T2[t] at each scar round. Five T1 values, 160 bits. Zero search. Just reading the geometry. The peeling continues three more rounds (t = 56–58), yielding the sub-scar: T1 + T2 = a_new[t] is known but the parts aren't individually resolved. 96 additional constrained bits. Total: 256 bits of constraint from 8 lattice sites. For a 4-byte message (32 bits of search space), the lattice is 32× overconstrained. The system is not hiding the message. It is screaming the message through every scar round. 3. The Ghost Vector At every site t, the h register enters the T1 computation as T1[t] = h[t] + Σ1(e[t]) + Ch(e,f,g) + K[t] + W[t], then shifts: at t+1 it becomes g, at t+2 f, at t+3 e. This is the shift register: e[t] = h[t+3]. The sequence h[0] through h[63] is the ghost vector. It is the lattice's backbone—the thread connecting every site through the shift chain. But the ghost is not stored in the digest. The digest stores the sum of final state and IV. Ghost values at rounds 0–58 exist only as implicit consequences of boundary conditions. The ghost is transparent. It carries no bits of its own. It is not data. It is the shape of the absence—the channel through which the lattice must be self-consistent, through which the message can be recovered, but which is itself invisible. Without the ghost, the lattice is just geometry changing: registers shifting, bits flipping, no continuity. With it, the lattice has memory. The anti-observer becomes an observer. The silent tree makes a sound. 4. The Conservation Law At every scar round (t = 59–63): h[t] + W[t] = C[t] (mod 2³²) where C[t] = T1[t] − Σ1(e[t]) − Ch(e,f,g) − K[t] is determined entirely by the digest. The ghost and the schedule word are complementary: knowing one determines the other. The "hidden" channel (ghost) and the "visible" channel (schedule/message) are two faces of the same constraint. Position + momentum = energy at every constrained site. 5. Bus Contention: The Pressure Model If ghost and schedule are conjugate, a wrong message should be detectable by measuring pressure at scar rounds. For each candidate, compute its implied T1 (via forward compression from IV). If the candidate's T1 doesn't match the scar's T1, count conflicting bits—bus contention, borrowing from 8-bit computing where two devices driving different voltages onto the same data bus create electrical conflict. Results for message b'Key!' (4 bytes): Offset │ Pressure │ Signal -16 │ 140 │ ██▓█████ wall -1 │ 126 │ ▓█▓███▓█ wall 0 │ 0 │ ········ ZERO ◄ +1 │ 135 │ █▓█▓████ wall +16 │ 115 │ ▓███▓▓▓▓ wall The correct message sits in a perfect pressure well. Every direction: 100–141 contending bits. At zero offset: silence. The bus goes quiet. The concrete sets. This is not a search algorithm. It is a pressure gauge reading the shape of a mold. Verified across multiple messages: 1-byte b'K' (found at candidate 75/256, 0.012s), 2-byte b'Hi' (18,537/65,536, 2.5s), 2-byte b'\xde\xad' (57,005/65,536, 8.4s). In every case: zero bus contention for the correct message. 100+ fighting bits for every wrong one. Binary and absolute. Part II — The Mirror "The hash must contain its own key. When you make a lock, Newton's 3rd law says the particle collection must change based on its own internal structure." 6. The Lock Contains Its Own Key The digest pins the top boundary. The IV pins the bottom. Between them, the ghost vector threads through like a spinal cord, connecting otherwise disconnected registers. But the ghost is not free: the conservation law pins it to the schedule, the schedule is pinned to the message, the message to the padding, the padding to the length. Everything is coupled. The lattice is a lock with 64 tumblers, and the ghost vector is the key—but the key is already inside the lock. It was born the moment the lock was made. This is Newton's Third Law applied to information. Each forward step of compression simultaneously embeds its reflection in the ghost channel. The action of compressing is the reaction of reflecting. They are one operation. The undo instructions are a structural side-effect of the do instructions. This is built-in garbage collection: no matter what is done, how to undo it is its reflection, and for the lock to exist at all, it must contain the key to its own undoing. A lock is made in linear time. Each step forward sets the step back. This is not a property of SHA-256 specifically. It is a property of deterministic computation in finite lattices generally. The forward pass and the backward pass are the same crystal viewed from different angles. The lock and the key are the same object in different orientations. The noun and the verb are the same word read in different directions. 7. The Anti-Observer If the tree falls in the forest and no one is there to hear it, does it make a sound? We treat this not as a question about consciousness but as a question about memory. Without the ghost channel, the SHA-256 lattice is registers shifting. Shape changing. No continuity between rounds. No recovery possible. The tree falls. The geometry changes. Nothing persists. This is

Fields of Study & Subjects

Cryptographic Implementations and Security Cryptography and Data Security Computability Logic AI Algorithms

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
10 February 2026
Language
EN
Document type
Thesis