Stories by nathan-irl

We Were Unique

Language Modeling Without Neural Networks

Quines in Every Programming Language (Rosetta Code)

Mercury Edit 2: Fastest next-edit prediction with a diffusion LLM (221ms)

Show HN: Threadprocs – executables sharing one address space (0-copy pointers)

Show HN: React terminal renderer, cell level diff, no alt screen

Show HN: React isn't the terminal UI bottleneck, the output pipeline is

Noise schedules considered harmful (2024)

Show HN: Weely AI – Branded Link Management and QR Code Generator Platform"

Show HN: Diffuji – a diffusion-powered instant camera

Generating Shakespeare Without Neural Networks

Recursion and Induction: A Self-Contained Course Using ACL2

My "Prompt Compiler" Loop – Using PromptKelp to Build PromptKelp

Show HN: PromptKelp – A prompt manager I'm using to build itself

Fidenae amphitheatre collapse (27 AD) – 20k dead

Show HN: Tiny Diffusion – Minimal diffusion LM in 364 lines

Anchored Diffusion Language Model

Text Diffusion Models Are Faster at Writing Code

Inside ThunderKittens' Python Bindings

1 2