gensyn-primary-sources-2026-05-14

Gensyn primary sources

Reviewed: 2026-05-14 UTC

Current docs framing

• https://docs.gensyn.ai/
  • Best current source for how Gensyn wants to be understood now.
  • Used to confirm the present framing of Gensyn as the Network for Machine Intelligence and an open infrastructure layer for compute, data, and information exchange.
  • Also useful for showing that the public docs now foreground live products and ecosystem/research surfaces rather than only a monolithic protocol spec.

Legacy protocol-design document

• https://docs.gensyn.ai/litepaper
  • Still the clearest canonical protocol-design document, but the page now explicitly warns that it is out of date.
  • Used to confirm the original problem statement: trustless verification of offchain deep-learning computation, along with the original participant roles (submitters, solvers, verifiers, whistleblowers) and the older L1-oriented design language.
  • Most important meta-signal from this page is the official change notice: Gensyn says it replaced the earlier Substrate L1 approach with a custom Ethereum rollup and has since added stronger audit strategy, cryptographic proof work, availability guarantees, and reproducible-runtime components.

Information-market product

• https://docs.delphi.fyi/
  • Best current source for the information exchange branch of the Gensyn stack.
  • Used to confirm that Delphi is an AI-settled information-market product where creators configure a model and settlement prompt, and where some markets support verifiable settlement via Gensyn REE receipts that can be independently rerun.
  • Helpful for showing that Gensyn is not just building compute rental or training middleware; it is also productizing AI-mediated market settlement.

Open-source training surface

• https://github.com/gensyn-ai/rl-swarm
  • Best practical source for the current public participation surface.
  • Used to confirm that RL Swarm is an open-source, permissionless peer-to-peer reinforcement-learning system, that the current active environment is CodeZero, and that node operators can participate with consumer or datacenter hardware.
  • Also useful for confirming that Gensyn exposes testnet-linked identity / participation flows rather than keeping experimentation entirely offchain.

Public repo map

• https://github.com/gensyn-ai
  • Useful high-level source for how the team’s public work is split across repos.
  • Confirms that RL Swarm, RL-swarm contracts, CodeAssist, BlockAssist, and research-code repos are all part of the public surface, reinforcing the interpretation that Gensyn is an ecosystem of training, execution, and application layers rather than a single narrow protocol binary.

Research posture

• https://www.gensyn.ai/research
  • Lightweight source, but useful for corroborating the project’s own emphasis on decentralized machine learning and algorithmic economics.
  • Supports the framing that the stack is still research-heavy and evolving.

Main takeaways from this pass

• Gensyn should not be filed only as a decentralized compute marketplace or only as a legacy trustless-training L1 concept.
• The current materials expose at least three comparison-ready layers: decentralized training swarms, reproducible / verifiable model execution, and AI-settled information markets.
• The out-of-date-litepaper warning is itself one of the most important source facts because it shows the official architecture has materially moved since the 2022 protocol sketch.
• Gensyn is therefore strongest in the corpus as a crypto × AI coordination stack where verification policy, participation design, and market structure are the main analytical control surfaces.