01Digital Hana Arts

MLabChain is developed and maintained by Digital Hana Arts, a small studio building honest, research-grade tools in the intersection of machine learning, scientific reproducibility, and distributed systems.

DHA's work is deliberately narrow. The studio is not trying to build a platform, a network, or an investment vehicle. It builds tools that solve a specific problem, states the problem's limits clearly, and releases the code under a permissive license so that anyone can use, study, or extend it.

Focus

Reproducibility

Tools that help researchers prove what they actually did, long after the fact.

Focus

Applied cryptography

Small systems that use hashing, signatures, and Merkle proofs for real, narrow problems.

Focus

Honest documentation

Every project states what it does, what it does not do, and where its trust boundaries are.

Focus

Local-first software

Tools that run on one machine, need no external service, and outlive their authors.

02Project history

MLabChain grew out of a much older project called HEPLabChain, which was a provenance ledger for high-energy-physics analyses. That project recorded the hashes of options files, ROOT tuples, and fit results so that an LHCb analyst could prove, months or years later, that a particular file had not been modified.

MLabChain generalises the same idea from physics to machine learning. The scientific layer was rewritten to train models instead of just hashing them. The chain layer was rewritten in C++ to make signatures, replay, and rewards auditable independently of the Python client. The reward formula went through several iterations, and the current marginal-against-best design was the result of trying to prevent duplicate payment without adding centralised challenge review.

v0.1.7
Python-only provenance ledger

SHA-256 hashing, Merkle trees, RSA-signed transactions, JSON chain, non-transferable credits.

v0.2
First C++ core, first Mera

Ed25519, encrypted wallet, SQLite state, integer reward arithmetic, transferable balances. Still single-node.

v0.3.0-devnet
Small-network reference protocol

Signed blocks, cumulative-work fork choice, adaptive PoW, challenge registry, marginal rewards, P2P sync and relay, multisig, governance, ERC-20 template.

Next
Correctness and coverage

Fixes for marginal reward estimation on the client side, expanded test coverage, and the bridge specification document.

03Credits

Protocol design, C++ core, Python client, and documentation: Ali Bavarchee.

The project draws on decades of prior work in cryptographic primitives, consensus algorithms, and Merkle tree design. The specific sources that shaped the current implementation are listed in the further reading section.

The mascot, Hana, was drawn by a member of Digital Hana Arts and has been part of the project from the beginning. She approves of the 60-second target block time and has no opinion on the retarget window.

04Hana

Hana the kitten

Hana

Hana is a kitten. She is the official mascot of MLabChain and reminds everyone that a ledger is only as useful as the analysis it makes reproducible.

She does not sign transactions. She does not appear in any transaction payload. She has never been observed in the vicinity of a Merkle root. She is, however, aware of the project's reward cap, and she considers it reasonable.

— Digital Hana Arts

05License

MLabChain is released under the Apache License 2.0. You are free to use, modify, and distribute the code, including for commercial purposes, provided that you retain the copyright notice and the license text.

What Apache 2.0 gives you

A permissive license with an explicit patent grant. It is suitable for adoption by institutions and by companies, and it does not require derivative works to be open-source. If your intent is that derivatives must remain open, you would want a copyleft license instead. Apache 2.0 is the right choice for a scientific tool where the goal is maximum adoption.

The full license text is available in the repository at LICENSE.txt.

06Contact

The project is developed on GitHub at digitalhanaarts.

Bug reports, feature requests, and pull requests are welcome through the standard GitHub workflows. For security disclosures, please open a private security advisory on the repository rather than a public issue.

The docs page describes the CLI and protocol in detail. The protocol page covers the architecture and the honest boundary.


Return to the home page →