xotiq

(pronounced "exotic")

The Hardware Description Language for Hybrid Physics-Based Computing.

xotiq is a language for designing systems where deterministic control (digital logic) interacts directly with probabilistic physics (analog dynamics).

It provides a unified abstraction for the emerging class of "post-Moore" accelerators (whether photonic, spintronic, or thermodynamically-driven) allowing engineers to describe computation as a process of energy minimization and physical settling, rather than just boolean switching.

---

The Problem: The "Digital-Analog" Gap

Modern frontier hardware is hybrid. We are building chips that combine standard digital logic with "exotic" physical substrates (photonic meshes, magnetic tunnel junctions, valleytronic gates) to solve optimization and inference problems.

Current tools force a choice:

• Verilog/VHDL: Perfect for the digital control logic, but cannot model the continuous physics of the accelerator fabric.
• SPICE/Multiphysics: Perfect for the physics, but cannot model the complex state machines required to drive them.

xotiq bridges this gap. It treats the physical fabric not as a "black box" peripheral, but as a first-class citizen in the logic graph, managed by a compiler that understands both clamping (control) and settling (inference).

---

The Abstraction

xotiq designs are "bipartite graphs". The compiler manages the impedance mismatch between two distinct domains of computing:

The Deterministic Domain (d_node)

• Role: The "Controller." Handles sequential logic, state machines, I/O, and data routing.
• Behavior: Classical, restorative, non-linear.
• Physical Realization:
  • Standard CMOS Logic (Gates, FSMs).
  • Photonic Resonators: Microring resonators or saturable absorbers that "clamp" signals to binary states.
  • Spintronics: Stable, high-barrier magnetic junctions.

The Probabilistic Domain (p_node)

• Role: The "Fabric." Handles pattern matching, optimization, and massive parallelism via physics.
• Behavior: Continuous, interferometric, energy-based.
• Physical Realization:
  • Photonic Meshes: Mach-Zehnder Interferometer arrays doing passive matrix math at the speed of light.
  • Stochastic Logic: PRNG-driven digital fabrics.
  • Nanomagnetism: Superparamagnetic islands interacting via dipole coupling.

The Bridge

Computation is defined by the interaction between these domains:

• clamp(node, value): The Digital domain forces a Physical node to a specific state (e.g., applying a voltage bias, tuning a laser input).
• observe(node): The Digital domain reads the settled state of the Physical fabric (e.g., ADC, Photodetector).

Zig "SDK" example code:

// A Hybrid Optical Logic Gate
const std = @import("std");
const xq = @import("xotiq");

// The Fabric (p_nodes)
// A passive mesh of waveguides and couplers defined at compile-time.
// In a photonic backend, this compiles to an interferometer mesh.
const XOR_Topology = struct {
    A: xq.Node,
    B: xq.Node,
    Output: xq.Node,

    // Define the physics: Destructive interference
    pub const links = .{
        .{ .src = .A, .dst = .Output, .weight = -1.0 },
        .{ .src = .B, .dst = .Output, .weight = -1.0 },
    };
};

// The Control (d_nodes)
// The "Active" components that drive the laser and read the result.
pub fn main() !void {
    // Synthesize the fabric
    var chip = try xq.synthesize(XOR_Topology);

    // "Clamp" inputs (Inject Coherent Light)
    chip.clamp(.A, 1.0);
    chip.clamp(.B, 0.0);

    // Wait for light to propagate (Physics happens here)
    chip.wait(10 * xq.ps);

    // "Observe" the result (Read the Photodetector)
    // Implicitly handles analog-to-digital thresholding
    const result = chip.observe(.Output);
}

---

Compiler Backends

The xotiq toolchain uses a multi-stage lowering strategy to target increasing levels of physical realism.

• Target: Digital Emulation (CPU/FPGA)

  • Output: SystemVerilog / C++. / etc.
  • Method: p_nodes are lowered to pseudo-random number generators and digital accumulators. Allows for rapid algorithm verification on standard hardware.

• Target: Analog Synthesis (FPAA/ASIC)

  • Output: SPICE Netlist.
  • Method: p_nodes map to op-amp summing junctions or resistor networks; d_nodes map to standard voltage-mode logic.

• Target: Integrated Photonics

  • Output: GDSII / Circuit Simulation Interconnects.
  • Method:
  • WDM Support: Topological links are assigned specific wavelengths ("colors"), enabling massive parallelism on single waveguides.
  • Non-Linearity: d_node logic is synthesized as resonator/SOA structures for signal restoration.
  • Interference: p_node meshes are synthesized as passive optical linear units.

---

Contributing

xotiq is an open effort to build the standard infrastructure for post-Von-Neumann computing. Core development focuses on the Intermediate Representation (IR) and the Zig-based reference compiler.

• Logic Designers: Help define standard libraries for probabilistic arithmetic.
• Physicists: Help refine the backend models for specific photonic and spintronic constraints.
• Compiler Engineers: Help optimize the lowering passes for hybrid FPGA/ASIC targets.

---

Inspiration & Citations

xotiq is synthesizing concepts from quantum mechanics, thermodynamics, and high-performance computing.

Foundations

• Simulating Physics with Computers (1981)

Richard Feynman. The foundational argument for why nature cannot be efficiently simulated by classical boolean logic.

• Minds, Brains, and Programs (1980)

Crucial for understanding the distinction between syntactic manipulation (traditional computing) and semantic understanding (physical settling).

Architecture & Hardware

• All-Optical Digital Computing Using a Single All-Optical Gate Type (2024)

The architectural proof-of-concept from Akhetonics, demonstrating how non-linear optical gates (SOAs) can form the universal "NAND/NOR" logic required for deterministic control.

• Extropic: An Efficient Probabilistic Hardware Architecture for Diffusion-like Models (2025)

Demonstrates the viability of energy-based probabilistic hardware for modern AI workloads.

• Massively Parallel Probabilistic Computing with Sparse Ising Machines (2021)

Validates the "Ising Machine" approach (a core p_node behavior) for solving combinatorial optimization.

• Anatomy of High-Performance Matrix Multiplication (2008)

The "Bible" of classical matrix math optimization; a reminder of the deterministic bottlenecks we aim to bypass.

• Field Free Spin-Orbit Torque Controlled Neuron Devices for Spintronic Boltzmann Neural Networks (2025)

A direct physical candidate for the spintronic implementation of p_nodes.

Theory & Logic

• Information Physics of Intelligence: Unifying Logical Depth and Entropy under Thermodynamic Constraints (2025)

Provides the thermodynamic framework for treating computation as energy minimization.

• Computing with Time: Microarchitectural Weird Machines (2024)

Explores the use of analog timing and "settling" as a computational primitive, directly relevant to xotiq's wait() and observe() semantics.

• Inconsistency Robustness in Logic Programming (2009)

Essential reading for designing control logic (d_nodes) that can tolerate the inherent noise of physical fabrics.

• Spacetime Hopfions from Skyrmion Braiding (2023)

Sometimes, (maybe always?), state is just topology.

---

© 02026 Wilson Bilkovich