Writing

Notes, write-ups, and project stories.

• When 503s Are the Right Answer (Part 2 of LLM Inference Bottlenecks)

  September 1, 2026

• What Filled Up First? Diagnosing LLM Inference Bottlenecks

  August 31, 2026

• Why My 4-Bit CUDA Kernel Wasn't Faster

  August 8, 2026