Xenomorph Blockchain

We believe the connection between blockchain infrastructure and useful computation deserves careful exploration. Our focus remains on technical development, testing and community feedback. Follow our progress and join the discussion with the Xenomorph community on Discord.

A key challenge is verification. A decentralized network must determine whether a submitted task was executed correctly and how it relates to the intended model update. This is an important area to follow as the UsefulPoW design develops.

Our longer-term direction is UsefulPoW: connecting distributed computation with AI model training. Models such as DNABERT-2 offer a relevant research context for genomic sequences, but useful training requires more than processing data. It needs verifiable results.

GenomePoW uses deterministic transformations on genomic data as part of its mining process. This gives our approach a different structure from conventional hash-based mining, but the algorithm's design and its security properties must always be evaluated separately.

Traditional Proof of Work secures a blockchain through computation. With Xenomorph, we're exploring how genomic data can play a role in mining through GenomePoW, while developing the broader idea of UsefulPoW. The goal is to make computation more meaningful.

This is the complete flow: you mine, you train, the model generates value, you receive a portion of that value. We're testing this on the testnet right now. If you want to follow the development in real time, Discord is where it all happens - discord.gg/frCdBRXHuP

Payments for model queries are automatically distributed: 70% to seed nodes, 20% to the miners who contributed to the training, and 10% to the treasury. The logic is in the InferencePayments contract, not in manual decisions.

Once trained, the model becomes available as a service. Anyone can access it by paying in USDT. It's not just a blockchain—it's AI infrastructure with a concrete use case that can be verified from the base protocol.

Miners run training tasks on DNABERT-2, a model specialized in genomic sequences. Each task is verified using ZK proofs—the network confirms that the work was performed correctly without repeating the entire calculation.

In traditional mining, the work ends when the block is found. In Xenomorph, mining is the beginning of something more: each executed task contributes to training a real AI model on human genome data.

Xenomorph combines BlockDAG and UsefulPoW as the foundation of its protocol. Although still in the testing phase, the direction is clear: a network where mining work does not go to waste. Real-time development updates are available on our Discord.

UsefulPoW requires continuous and verifiable computation. An architecture that minimizes discarded blocks allows more training tasks to reach consensus. That is part of what makes this combination technically sound.

In traditional mining, if two miners find a block at the same time, one is discarded. In Xenomorph with BlockDAG, both blocks can be valid. This results in less wasted work—and with UsefulPoW, every training task executed counts.

In a linear blockchain, blocks compete with one another, and only one wins. In BlockDAG, multiple valid blocks coexist in parallel. Processing more blocks simultaneously means more opportunities for useful work to occur continuously.

The result: a mining algorithm that any GPU can run competitively, preventing specialized hardware manufacturers from centralizing the network. ASIC resistance by design, not by configuration. Public code available on GitHub.

ASICs dominate Bitcoin because they are optimized for fixed, repetitive calculations. Genome PoW changes the memory access pattern with every iteration by utilizing the genome data. An ASIC designed for this task holds no real advantage over a GPU with sufficient VRAM.

Why is it "memory-hard"? Because solving the puzzle requires loading and querying large blocks of genomic data in RAM. Having fast chips isn't enough—you need actual memory. This makes the algorithm resistant to specialized ASICs.

GRCh38 is the reference map of the human genome, comprising approximately 3.2 billion base pairs. Genome PoW extracts fragments from that sequence and applies deterministic mutations to generate the puzzles that miners must solve.

Our mining algorithm is called Genome PoW. It doesn't use arbitrary hashes; instead, it uses actual fragments of the human reference genome (GRCh38) as the basis for its calculations. Every mining operation works with real biological data.