We don't only buildon the network.We run our own.

Every turn, a coding agent re-sends the whole conversation - the files it already read, the tool output it already used, its own reasoning. You pay for all of it. SOMA compresses the context and the chain-of-thought on the way to the model, and keeps what the task needs.

The compressors are not ours alone. Subnet 114 runs them as an open competition: independent miners submit, validators score every submission on real tasks - cost and whether the task still passed - and after each round our team takes the winning code and makes it production-ready. Usage funds the next round.