/ about

coral.about

a distributed evolutionary compute network — open-source, decentralized, and built for researchers who want to see what a mesh of machines can evolve on its own.

scroll

/ about

what is coral

coral is an open-source distributed evolutionary compute network. it is a research platform, an engineering experiment, and a living demonstration of what a decentralized mesh of machines can produce when given a shared goal and left to run.

the network is composed of miner nodes — node.js processes that anyone can run. miners evolve experimental cryptographic transformation structures: substitution tables, permutation layers, mixing matrices, round constants. they benchmark their candidates, propagate the best results to peers via a gossip protocol, and vote at epoch boundaries on which candidate survives into the next generation. the network never stops. it evolves continuously.

once per hour, the scheduled pipeline chain collects the network's latest output and publishes a finalized snapshot to a public github archive. every epoch winner, every lineage relationship, every benchmark score — permanently recorded, permanently public.

coral is not a cryptocurrency. there are no coins, no tokens with financial value, no trading, no speculation. contribution artifacts — soulbound badges that prove a miner's participation in a given epoch — are the only token-like objects in the system, and they are explicitly non-transferable and worth nothing financially. the only thing they prove is that you ran a node and contributed compute.

coral was designed and built by sai satvik. the core design principles: separation of responsibilities between layers, resilience at every level, lightweight coordination that does not become a single point of failure, and a public archive that outlives any individual component. the miner network can continue operating even if the cloudflare worker goes offline. the github archive is public and permanent regardless of what happens to the rest of the infrastructure.

the goal is not production-grade cryptography. the structures coral evolves are not ready for, and should not be used in, real systems. the goal is to demonstrate that a distributed network of independent nodes, guided only by a shared fitness function and a lightweight consensus mechanism, can collaboratively produce and refine complex structured outputs over time. that is interesting on its own terms.

coral is built in the open. all source code is public. all evolution output is public. the protocol is documented, the architecture is documented, and the decisions behind every design choice are explained. if you want to understand how it works, you can read every line. if you want to run a node, you can be online in ten minutes. if you want to improve it, pull requests are welcome.

the network is small. the decisions made now will shape what coral becomes. individual contributors matter. if you find this interesting, run a node.