Category: crypto × AI training infrastructure / information-market and verifiable-execution stack / decentralized machine-intelligence coordination layer
Summary: Gensyn is best understood not as a single fixed decentralized GPU marketplace or the frozen layer-1 design described in its 2022 litepaper, but as an evolving crypto × AI coordination stack that now combines open machine-training infrastructure, reproducible execution, and AI-settled information markets. The current docs frame Gensyn as a network for machine intelligence built around open markets for compute, data, and information exchange, while the legacy litepaper still matters because it shows the original trustless-ML verification problem the project is trying to solve. The most reusable mechanism insight in the current materials is the split between training swarms such as RL Swarm / CodeZero, reproducible execution and receipt-based verification for model runs, and Delphi-style information markets where AI models can act as settlement logic. That makes Gensyn a useful comparison point for crypto × AI systems where the real control surfaces are verification policy, market design, training participation, and model-execution reproducibility rather than just raw compute rental.
What it does:
Frames itself as open infrastructure for machine intelligence, with markets for compute, data, and information exchange
Publishes research and experimental systems for decentralized machine learning, including scalable verification and communication frameworks
Operates or documents training environments such as RL Swarm / CodeZero, where distributed participants contribute compute to collaborative reinforcement-learning tasks
Exposes an AI-settled information-market product, Delphi, where creators configure markets and choose model-driven settlement logic
Uses reproducible-execution and receipt concepts so some AI settlements can be independently rerun and checked rather than trusted purely as API outputs
Maintains open-source repos spanning RL training swarms, onchain coordination contracts, and adjacent AI-agent experimentation
Key claims:
The current docs explicitly say Gensyn is the Network for Machine Intelligence and describe three market surfaces — compute, data, and information exchange — which is a broader framing than the older trustless deep-learning L1 pitch.
The legacy litepaper is still useful because it states the original hard problem clearly: verifying offchain deep-learning work without naive full replication. But the page now warns that the litepaper is out of date and lists major design changes, including replacing the original Substrate L1 approach with a custom Ethereum rollup and adding stronger audit, proof, and reproducible-runtime work.
Delphi matters because it shows Gensyn is not only building training infrastructure. The official Delphi docs frame information markets as a live product where creators choose AI models and settlement prompts, and where some markets can use reproducible execution receipts for independently checkable settlement.
RL Swarm is important because it makes the participation surface legible. The current repository describes an open, permissionless peer-to-peer reinforcement-learning system that ordinary users can run on consumer hardware or GPUs, with onchain identity tracking for testnet participation.
The repo and docs surface together suggest Gensyn is best cataloged as a moving coordination stack rather than a finished monolithic protocol. Research, open-source training clients, onchain contracts, reproducible execution, and information markets all appear as distinct but related layers.
Gensyn therefore belongs in the active corpus as a comparison point for decentralized training networks, crypto × AI execution systems, and agent-market designs where verification and market structure matter as much as hardware aggregation.
Whitepaper: The official litepaper remains the clearest canonical protocol-design document, but the docs explicitly label it legacy and note major design changes since publication. The strongest primary materials in this pass were the live docs home page, the legacy litepaper, the Delphi docs, the RL Swarm repo, and the public GitHub organization; see ../whitepapers/gensyn-primary-sources-2026-05-14.md.