Using Antigravity for Statistical Physics in JavaScript (christopherkrapu.com)

The author used Google’s new Antigravity IDE with Gemini 3 Pro to generate a JavaScript visualization of the Ising model in about an hour. The implementation simulates a 2D lattice of spins (±1) and updates them with a Random Scan Gibbs Sampler: at each step a spin is chosen uniformly at random and resampled from the Boltzmann distribution determined by its nearest neighbors and an optional external field. The result is an interactive browser visualization of ferromagnetic behavior derived from the classic Ising formulation (Ernst Ising’s 1924 problem), produced with LLM-generated code augmented by a Chrome extension that reads and manipulates the DOM. This demonstration is significant because it shifts LLM benchmarks from factual recall toward rapid, practical prototyping of domain-specific simulations and visualizations. It showcases how modern models plus IDE tooling can produce nontrivial scientific code quickly, making computational physics demos more accessible and enabling iterative experimentation. Technical caveats include reliance on the DOM-retrieval extension to wire the UI and some IDE UI features (subtasks/intermediate-step displays) that the author found superfluous. Overall, the experiment highlights growing utility of LLMs for hands-on scientific coding and interactive education, while underscoring the role of dev tooling in turning generated code into working apps.