SlothDB is a super fast embedded SQL database
1.3MB WASM database beats DuckDB 5x on 10M row analytics, runs in Cloudflare Workers.
SQLite queries in Brainfuck via raw bytes and Linux pipes, no C helpers.
Language designers, constraint enthusiasts, AI/automation researchers, esoteric programming hobbyists
The idea is to see how far a deliberately minimal language can be pushed using only core Linux primitives (pipes, dd, file operations) and direct interaction with SQLite database files.
Brainfuck was chosen because it is intentionally extreme. Even compiling the generated code turned into an adventure: the Brainfuck compiler emits C, and the output was so large that gcc couldn't compile it, though clang could.
This is the sort of project no sane developer would normally attempt. With AI assistance, however, it became feasible to explore in part of a Saturday.
1.3MB WASM database beats DuckDB 5x on 10M row analytics, runs in Cloudflare Workers.
SQL queries on CSV streams—instant, zero-setup alternative to awk and sqlite3 boilerplate.
Click-to-expand DB schema instead of joins, but DBeaver, DataGrip, and SQuirrel do this already.
Finally searches iMessage properly by reading the database Apple won't expose.
Finally gives Claude Code persistent log memory without manual copy-pasting.
Runs real SQL JOINs on DynamoDB by auto-selecting indexes, skipping the usual ETL pipeline.