When thousands of requests share the same system prompt, recomputing its KV cache each time is pure waste. Prefix caching stores and reuses those vectors, cutting TTFT by up to 97% in common deployments.

Splitting a long prefill across multiple iterations keeps decode requests from stalling, with no extra FLOPs and negligible IO overhead.

How iteration-level scheduling eliminates GPU idle time, and how prefill and decode rows can share one packed forward pass.

How vLLM borrows the OS paging idea to eliminate KV cache memory fragmentation, pushing GPU utilization from ~30% to ~96%.

how online softmax extends the fused kernel to handle rows that exceed sram capacity, using a numerically stable 2-pass tiling algorithm.

why the key-value cache avoids redundant computation in autoregressive decoding, and the memory/compute tradeoffs it introduces.

how to write a fused softmax kernel in triton that eliminates redundant memory accesses and outperforms pytorch's native implementation.

A practical guide to SSH local and remote port forwarding, with examples, comparison, and persistent configuration via ~/.ssh/config.

Use mitmproxy to capture LLM API traffic and Tampermonkey to turn mitmweb's raw JSON into a readable, chat-like viewer.

understand batch gradient descent, stochastic gradient descent, and mini-batch gradient descent.

understand how backward propagation works in gradient descent.

learn how to manage you keys with GPG, and use it with ssh and git and pass.

DSU on tree answers subtree queries by keeping the largest child's contribution and rebuilding only the small parts. The trick is not union-find; it is small-to-large merging hidden inside a DFS.

compare shortest path algorithms: dijkstra, floyd, bellman-ford

this is a description

connect to android via TCP/IP

this is my i3 window manager configuration

Pros and cons of different types of cloud.

introduce randomness into your C++ programs, the right way.

manage passwords on all your devices (add, generate, edit, delete, sync).