Nova is a home for low-level engine work and hands-on AI experimentation: Nova_A builds a 2D engine/editor stack in C++20; Nova_v1 explores LLM training and chat in Python.
The public repository describes Nova_A as a fully open-source 2D physics engine, renderer and GUI editor built from scratch with C++20. Its current public README documents OpenGL/GLAD initialization, GLFW windowing, ImGui editor integration and a rendering abstraction.
Nova_v1 is described by its repository as a simple LLM training and chat application based on Python. This page intentionally keeps the description conservative and points to the repository for the exact implementation and current capabilities.
Keep the training workflow understandable: inputs, configuration, model state and outputs should remain inspectable rather than hidden behind an opaque service.
A chat surface closes the loop between model work and direct interaction, making it easier to test behavior and iterate on training experiments.
Nova_A works downward toward graphics, physics and editor primitives. Nova_v1 works upward toward model training and interaction. Both are useful because the implementation remains close enough to inspect, change and learn from.
Prefer explicit systems and readable boundaries over hidden behavior.
Keep versions testable so every change can be verified and rolled back.
Use the repositories as the source of truth for implementation details and progress.