Michael SimkinIndependent researcher · Tbilisi, GeorgiaSearch

Programming & Computing

GPU integration, search infrastructure, and the code that makes experiments possible.

Explore the 6 articles ↓

The implementation can be a contribution in its own right. This interest follows the engineering behind the work: combining GPU chess codebases in Deep-Formix, building LifeAPI and CatForce for combinatorial searches, and implementing adjustable-precision GPU arithmetic for a fractal explorer. The same articles also belong to their subject categories.

Articles in this topic

6 articles
A syntax-colored LifeAPI code example initializing and evolving a glider.
ProjectImplemented

LifeAPI

Optimized 64×64, bit-packed Game of Life evolution, with the operations needed for combinatorial searches.

Animated enumeration of different block and beehive catalyst placements.
ProjectImplemented

CatForce

Brute-force catalyst searches that test recovery without prescribing the interaction.

Conceptual illustration of a chess knight on a GPU chip with parallel circuit paths.
ProjectImplemented

Deep-Formix

A custom deep, narrow evaluation network, NNUE-style updates, and a working GPU chess search—developed while stress-testing Gemini.

Computed Mandelbrot set with colored escape-time bands.
ProjectImplemented

Mandelbrot CUDA

Deep fractal exploration using adjustable fixed-point precision on CUDA and WebGPU.

Both ends of Geminoid Particles build their next heads, then remove old circuitry, while the long instruction tape winds between them.
ResearchDemonstrated result

Geminoid Particles

A custom nine-state cellular automaton for exploring construction and self-destruction.