7-Zip Compression Speed Boosted by 97% Without Touching Core Code: An "Amateur" and AI's Extreme Optimization Experiment

Yusheng Science, an AI incubator, completed an optimization of 7-Zip with an employee who had no programming background. The idea was to improve compression efficiency by using a "content-aware router" that dynamically selects encoding strategies based on file type, instead of using a uniform strategy for all files. The optimization involved a multi-level detection mechanism (file header recognition, sample compression testing, and multi-point sampling for suspected compressed media), a ten-tier routing strategy, and safety measures (sample comparisons, tolerance thresholds, and grouping isolation). Results showed a 97.0% compression speed increase for multi-file archives on macOS, 93.4% for prose text, over 80% typical on Windows, and a 40.4% size reduction for logs, with 100% integrity verification. However, decompression time for text and log files increased 2-3 times, and very large incompressible files over 5GB may suffer slower compression due to misjudgments. The optimization is zero-invasive: no core algorithm code was changed, only the dispatch layer, and output remains standard .7z format, decodable by official 7-Zip. The optimized version is open-sourced on GitHub (https://github.com/sunmmer01/Optimization). Yusheng Science states this practice demonstrates a new collaboration model where AI amplifies human insight, not replacing engineers.