Genseki is the main quest: a Hive engine lab where neural research has to survive real rules, real opponents, and evidence I can reproduce.
01 Current priority queue
Four projects, in order. Genseki comes first; Eloi stays paused until its Kaggle program has an owner.
#1I saw a small bot ecosystem for a big-player-base board game and decided it was mine. Genseki is my C++26/Python Hive engine lab.
#2My Super Mario 64 DS PC-port lane, kept in sync with the credited decompilation and co-op foundations it builds on.
#3I’m building an experimental C++26 route toward running Mario Kart 7 natively on PCs and laptops instead of hiding everything behind full-system emulation.
#4 / pausedMy chess-engine rabbit hole has a Rust 2024 four-player branch, but I’m not actively working on it until somebody takes on the v4.0.0 Kaggle program for me.
completeI turned Life Engine into a native evolutionary ecosystem where organisms inherit bodies, physiology, perception, and sparse neural behavior.
.rwI made a text-based language for learning programming without hiding all the real structure.
I’m seeing what a dynamically typed, JIT-compiled language for the web can feel like.
.pyI wrote a Python interpreter in C and C++ because apparently using Python normally was too easy.
01B GitHub detour
A blunt future goal: recompile Super Mario 3D Land for native hardware. The repository is a marker, not fake progress.
A recent experiment I keep around because “plan first” is still a technical result.
A deterministic top-down racer with a recurrent neural driver. I learned from it; I am not pretending it is still active.
A configuration-language experiment that evaluated trusted source into deterministic JSON.
A fast terrain generator with biomes, erosion, rivers, zoom, and pan in one Python file.
The deliberately small builds I use to keep trying ideas without pretending each one is a startup.
My earlier C++23 and Vulkan web-layout renderer experiment. Finished as a chapter, kept as evidence.
My archived attempt to rethink a few parts of Julia.
I built an open-source multiplayer Svelte game around collaboration.
A tiny interpreter for a language I made up.
What I built during OpenAI Build Week 2026.
My Hack for Humanity 2026 project about mental wellbeing.
What I shipped for United Hacks V7.
My open collection of custom swordbattle.io mods.
My multiplayer sword-fighting playground for quick experiments.
02 Neural work, zoomed in
Rho is my policy/value track: rules-grounded state, self-play evidence, training data, and experiments that stay separate from the Alpha engine.
I care about controlled opponents, frozen inputs, reproducible runs, honest failures, and knowing exactly why a model was promoted.
The model still has to meet a rules engine, search, a GUI, performance limits, packaging, and users. I like owning that entire path.
03 A little about me
I started with Python, got curious about what was underneath it, and ended up in C++ and Rust building game engines, neural agents, native ports, and the infrastructure around them.
I will apply to Google DeepMind. Until then, I’m turning the gap between curiosity and research engineering into public, reproducible work.
Neural
agents
Engine
systems
Native
ports
Evidence &
reproducibility