Recursor
Recursor - Advanced Compression
Introducing Recursor - a next-generation compression engine
Recursor — High-Structure "Middle-Out" Compression Engine
A next-generation, structure-aware compressor that exploits patterns across diverse data formats. Understand structure FIRST, then compress.
What Makes Recursor Different?#
Traditional compressors (gzip, zstd, brotli) treat all data as random bytes. Recursor understands the semantic structure of your data:
- CSV files: Content-defined chunking with row-group deduplication
- Code: Template extraction, pattern recognition across files
- Images: Block-based deduplication, palette quantization
- Audio: LPC (Linear Predictive Coding) + Rice coding + stereo decorrelation
The result: lossless compression that beats industry standards by exploiting high-level patterns.
Proven Results#
| Data Type | Size | Baseline | Recursor | Improvement |
|---|---|---|---|---|
| CSV (1M rows) | 174 MB | 60 MB (zstd-19) | 32 MB | 5.18x better |
| Code (36 JS files) | 144 KB | 3.3 KB (tar.gz) | 611 bytes | 16.9x better |
| Images (patterns) | - | 4,513 bytes (PNG) | 425 bytes | 10.6x better |
| Audio (synthetic) | 882 KB | 166 KB (FLAC -8) | 140 KB | 18.4% better than FLAC |
All Recursor results are 100% lossless with perfect round-trip reconstruction.