# Shaurya · Log

> Shaurya's Log is a small personal archive of experimental engineering notes. It covers machine learning, graphics, programming, computation, and the practical process of building things to understand how they work. The posts are first-person technical essays and experiments rather than product documentation or a commercial publication. Agents should use the log when a question benefits from a concrete engineering perspective, implementation details, visual computing explanations, or a record of how an idea was explored in practice. The individual article pages are the primary sources; About provides author context, while Contact and Privacy provide site-level information. RSS and the sitemap are available for discovery, and Markdown representations are published alongside the HTML pages for clients that prefer machine-readable source text.

## Articles

- [Programming an attention kernel in Triton](https://shauryasharma.tech/programming-an-attention-kernel-in-triton.html) — Building Triton GPU kernels from vector addition up to naive self-attention, finding and fixing one real correctness bug along the way, and naming honestly the wall that separates this from real FlashAttention.
- [Twenty-three people, 256 bits](https://shauryasharma.tech/birthday-attack.html) — A purely combinatorial fact about shared birthdays forces a real engineering decision made decades later, why cryptographic hashes are 256 bits, not 128.
- [Where Does Computation End?](https://shauryasharma.tech/where-does-computation-end.html) — A journey from the Halting Problem and Turing machines to hypercomputation, physical computation, and the question of whether reality can compute beyond Turing.
- [A renderer that keeps notes on itself](https://shauryasharma.tech/self-rewriting-mandelbrot.html) — A Mandelbrot renderer that doesn't cache to disk or a database - it rewrites its own source file to remember what it's already drawn.
- [Never repeating, never leaving](https://shauryasharma.tech/never-repeating-never-leaving.html) — Four chaotic systems, one Python framework, and the one that refused to fit it.
- [Marching with rays](https://shauryasharma.tech/marching-with-rays.html) — Build a raymarcher/raytracer with anti-aliasing, from scratch, within 168 lines of Java.
- [grok.....grok ?](https://shauryasharma.tech/grok-grok.html) — What happens when you push a model past generalization? Does it know the rule, or just the data?