AI agent infra stack map: frameworks, compute, wallets, privacy, inference, data
Post181 words
Maps the infra categories key for AI agents with example projects: agent frameworks, decentralized GPU networks, TEE and MPC agent wallets, chain-abstracted smart wallets, privacy-preserving compute, verifiable inference and programmable data feeds.
Infra that already is and will increasingly will be key for AI agents:
- Agentic frameworks and platforms that enable the creation of autonomous social & onchain agents.
> Study @ai16zdao (eliza), @virtuals_io, @EmpyrealSDK, @TalusNetwork, @autonolas, @Spectral_Labs, +++
- Decentralized GPU networks and liquid compute markets for agents to tap into and buy their own compute resources.
> Study @ionet, @AethirCloud, +++
- TEE-based key generation and MPC-based signing solutions with sound guardrail policies for agent wallets.
> Study @LitProtocol, @PhalaNetwork, @coinbase CDP, @biconomy +++
- Chain-abstracted smart wallets that agents can leverage to abstract the complexity of a fragmented multichain backend.
> Study @NEARProtocol, @ParticleNtwrk, +++
- Privacy-preserving compute infra to enable private LLM training and inference, unlocking use cases involving high-value data.
> Study @nillion. @arcium, +++
- Verifiable/decentralized inference layers. Includes optimistic ML solutions, zkML and crypto-economically secured inference layers.
> Study @ritualnet, @Gaianet_AI, @hyperbolic_labs, @AlloraNetwork, @gizatechxyz, @OraProtocol, +++
- Customizable, programmable and verifiable data feeds that empower truly autonomous, agentic decision-making at scale.
> Study @eoracle_network, +++
List not complete, there is more.
Just sharing some thoughts.