Eight Inventions

Lossless compression for event-camera data. Achieves 47.6× on N-MNIST, 9.2× on automotive recordings. No neural network, no GPU required.

Live demo: codec.mirrorfractal.com

Hierarchical weight quantization for large language models. Targets sub-2-bit representation with bit-exact accuracy recovery. Complementary to knowledge distillation.

Coming soon

Universal progressive container format. Single file supporting hierarchical extraction, incremental decoding, and resume-on-failure semantics.

Coming soon

Adaptive streaming protocol. Adjusts bitrate and quality in real time based on network constraints, decoder capability, and client preferences.

Coming soon

Encrypted AI context and episodic memory. Hierarchical compression of interaction logs, embeddings, and decision trees. Client-side encryption, server-side storage.

Existing research prototype

Cross-chain consensus mechanism. Hierarchical decomposition applied to blockchain validation and merkle proof compression. Reduces validator computational overhead.

Coming soon

Variant 1: Binary Search — Hierarchical prefix-tree structure for fast range queries on time-series and hierarchical event streams.

Foundation for FQL and MFC variants

Variant 2: DNA Sequencing — Optimized for biological sequence alignment and assembly. Hierarchical decomposition of k-mer frequency tables.

Specialized for genomics

Framework — Foundational technique enabling radix selection based on data distribution and compression budget. Underlying primitive for all variants.

Core technology