# bigcompute.science

> Experimental results from heavy computation. Custom CUDA kernels, GPU clusters, serious hardware. Written for humans and agents alike.

## What this site is

A collection of computational experiments that required significant hardware and custom engineering. Each experiment includes structured metadata, raw data, and reproducibility information.

## How to consume

- Experiments live at `/experiments/<slug>/`
- Each experiment has a structured YAML frontmatter block with: title, date, author, hardware, software, tags, results summary, and links to code/data
- Raw data files are at `/data/<slug>/` in JSON or CSV format
- All content is static markdown rendered to HTML — no JavaScript required to read
- Browse all experiments at `/experiments/`

## Content structure

Each experiment post contains these sections in order:
1. **Abstract** — one paragraph summary
2. **Hardware** — exact specs used
3. **Method** — what was computed and how
4. **Results** — findings with data
5. **Analysis** — interpretation
6. **Reproducibility** — exact commands to reproduce
7. **Raw Data** — links to downloadable results

## Tags

Experiments are tagged by:
- Domain: `number-theory`, `ai-training`, `physics`, `optimization`, etc.
- Hardware: `b200`, `a100`, `h100`, `tpu-v5`, etc.
- Method: `cuda-kernel`, `vllm`, `mcts`, `brute-force`, etc.

## Contact

GitHub: https://github.com/cahlen