Condense-json 1.0 released for compressing language model logs

Simon Willison released version 1.0 of condense-json, an open-source library developed over eighteen months to compress JSON storage in large language model logs. The tool scans JSON objects for repeated strings and substrings, replacing them with special {"$r": ...} syntax. This significantly reduces file size while maintaining complete data integrity. The compression is fully reversible using the uncondense_json function. Willison deployed the library to compress SQLite logs from LLM operations, achieving practical storage savings for production systems.