Tau: corpus representation, not compression

General-purpose compressors are blind to meaning. They see bytes, repetition, and entropy — and on a single Ethereum transaction, that leaves almost nothing on the table: isolated transactions compress only marginally. The opportunity is not inside one transaction. It is across them.

A blockchain transaction stream is not random. The same contract calls, the same calldata shapes, the same recipients and selectors recur across thousands of transactions and across time. Tau is built around that observation. It is a corpus representation engine: it models the structure of transaction history, not the bytes of a single payload. That is why an isolated transaction barely moves while a batched corpus reaches several-fold reduction — the gain is a cross-transaction effect, and it is honest about that.

For infrastructure buyers, the point is not smaller files. Sequencers, indexers, analytics, and archival systems care whether compressed data can be verified, streamed, queried, and reconstructed without operational uncertainty. So Tau is lossless and byte-exact on recovery, carries a constant-size inclusion witness, and supports querying without full decompression. It conforms to the public Ethereum test vectors byte-for-byte.

What Tau is not: it is not a universal compressor, not a blob replacement, not a Merkle replacement. It does not ask the market to believe — it publishes reproducible results and invites verification on your own corpus. The engine stays sealed; the evidence does not.