2026
- Online Softmax: Tiling for Arbitrarily Large Rows
- Why KV Cache Works in LLM Inference
- Fused Softmax in Triton
- SSH Port Forwarding: Local and Remote Tunnels Explained
- Mitmproxy + Tampermonkey = better {llm, …} viewer
- Batch vs Stochastic Gradient Descent
- Forward & Backward Propagation