A publication by Yash Datta

Read the Source.

Long-form, source-driven walkthroughs of the algorithms, numerical tricks, and engineering primitives that load-bear modern machine learning. New pieces by email at saucam.substack.com.

2026

  1. Jul 24 Long sessions without compaction AI agents
  2. Jun 23 Codeoid: cross-session memory and identity for Claude Code AI agents
  3. Jun 16 Three layers above your coding agents Agent security
  4. Jun 10 The rotation that rescues a 2-bit KV cache LLM inference
  5. May 26 Deploying MCP servers in production: the 2026 attack surface and the defense stack Agent security
  6. May 22 Whispers of Wealth, measured: 20 runs against the AP2 attack paper, on its own model Agent security
  7. May 21 What attention stores, and what DeepSeek changed Architecture
  8. May 20 MRV2: how vLLM rewrote its model runner, in six architectural moves LLM inference
  9. May 15 Streaming softmax: the recurrence that makes FlashAttention work Math foundations