Pinned Loading

1. TokenBurn TokenBurn Public

   TokenBurn: Information-theoretic metrics for measuring LLM confidence, hallucination risk, and optimal reasoning length.

   Python 18

2. CUDA-From-Scratch CUDA-From-Scratch Public

   A self-study for CUDA principles from 0 to 1

   Cuda 15

3. GPT-From-Scratch GPT-From-Scratch Public

   Transformer Architecture from first principles

   HTML 10

4. AutoKernel AutoKernel Public

   A production-quality terminal-based profiling tool that diagnoses GPU performance bottlenecks in PyTorch models.

   Python 8 1

5. Manimbot Manimbot Public

   Generate beautiful Manim animations using AI. Describe what you want to animate, and the system generates the code and renders it.

   Python 7

6. KVBoost KVBoost Public

   Make LLM inference faster with chunk-level KV cache reuse, prompt stitching, and zero-loss recomputation.

   Python