The csvql blog

Notes on building an analytical CSV query engine for AI agents.

2026-07-30

The Token Economics of Querying a CSV Instead of Pasting It

Pasting an 8 GB CSV into an LLM costs 230 million tokens and still doesn't fit. Querying it costs about 560 tokens, flat, regardless of file size.

2026-07-30

Why csvql joins beat DuckDB by 50x while aggregates only beat it by 8x

Two query types, two different bottlenecks. What the parallel join probe actually changed, and why the same trick doesn't move the needle on a plain COUNT(*).

2026-07-13

I Benchmarked My CSV Engine Against DuckDB on Its Own Dataset. It Found Two Bugs in Mine.

csvql queries raw CSV about 2.8x faster than DuckDB on an 8 GB file, using about 6x less memory and zero extra disk. The benchmark that proved it also surfaced two real bugs.