This post explores a simple idea: treating compression as an online optimization process, where mode switching is driven by how the compressor itself responds to the data, rather than by inspecting or classifying the data.
I implemented a small experimental controller applying this principle to existing compressors as a discussion artifact here: https://github.com/Alex256-core/AdaptiveZip
The goal is not a new compression algorithm, but a mathematically motivated way to select and switch modes automatically. Critical feedback is very welcome.