How GPT Works
Tokenization, embeddings, attention, transformer blocks, sampling, and training, ending with a tiny GPT-2-style decoder in JavaScript.
Filesystems and B-Trees
A practical guide to why B-trees matter for disk-backed systems, including search, splitting, deletion, and a miniature filesystem built on top of a B-tree.
Ropes and Piece Tables
The text-buffer structures behind modern editors, covering gap buffers, ropes, piece tables, undo history, and line-oriented access.
The Compression Game
Shannon entropy, Huffman coding, LZ77 back-references, and the way gzip combines both into a practical lossless compressor.
The Balanced Bookshelf
AVL trees, red-black trees, ordered maps, and the situations where a balanced tree is a better fit than a hash table.
The Casino Data Structure
Skip lists, probabilistic balancing, Redis sorted sets, rank queries, and a full skip-list implementation in JavaScript.
The Write-Heavy Problem
LSM trees, sequential write paths, MemTables, SSTables, compaction, Bloom filters, and the tradeoffs behind LevelDB, RocksDB, and Cassandra.
Good Enough Answers
Bloom filters, count-min sketch, HyperLogLog, and the engineering case for approximate answers when exactness is too expensive.
The Practical Champion
Quicksort, partitioning, pivot choice, introsort, heapsort fallback, and the cases where the fastest in-memory sorter is not the right tool.
Sorting Bigger Than RAM
External mergesort, run generation, k-way merging, sequential I/O, and a simulated BinaryFile-based external sort in JavaScript.