跳转到主要内容
image would be good

​
Livepeer 思维模型

Livepeer 是一种加密经济协作协议，确保全球按需 GPU 网络的安全，该网络针对实时视频和 AI 进行优化，并通过面向开发者的平台和应用程序进行公开。 你可以将 Livepeer 视为一个去中心化的无服务器 GPU 结构，拥有加密经济控制平面，服务通过一系列面向开发者的产品和应用程序进行公开。 Livepeer 类似于你访问和构建传统云服务提供商的方式，例如，社交媒体依赖于底层基础设施，你通过一套产品和应用程序（如 Twitter、Instagram、Facebook）进行访问。 Livepeer 的加密经济原语和去中心化计算网格为系统提供了额外的优势，例如抗审查、经济安全以及无信任协作。
The decentralisation and cryptoeconomic primitives of Livepeer add additional benefits to the system including
  • dynamic scaling & resilience (no single point of failure)
  • global payment rails (accessible by AI agents)
  • permissionless global participation & access (accelerates innovation and incentivises participation)
  • fairer creator economics (gives creators transparent, guaranteed revenue streams)
  • lower latency and data ownership (data stays local to the user)
  • lower costs without lock‑in (no middlemen)
  • incentives provide a runway to self‑sustaining, permissionless, global infrastructure without the need for venture capital (ramp to the tipping point)
  • network effects (as more participants join, the network becomes more valuable to all participants)
  • and more…

​
基础设施层

协议
A crypto‑economic protocol that secures and coordinates the network. (incentives, staking, governance, payments)
  • the protocol provides trust, coordination and payment mechanisms
Livepeer Protocol
网络
A global GPU network that performs the actual compute work. (supply, routing, performance)
  • the network supplies compute, routing, and verification
Orchestrators Gateways
应用程序和平台
A set of developer‑facing tools and applications (APIs, SDKs, apps) which expose the network’s capabilities in a usable way.
  • the platforms expose the network’s capabilities in a usable way.
Gateways Platforms

​
Livepeer 作为类似 OSI 的堆栈

映射基础设施架构的传统方法之一是使用 OSI 模型. 虽然该模型并不能完美映射到去中心化系统——且正被 AI 快速改造，但该框架仍有助于理解不同层次及其相互作用。 Livepeer 大致可以被视为一个 OSI 堆栈，其中“cloud control plane”被加密经济协议取代，持久存储是可选的、外部的，并且次于实时流式计算。

​
网络与协议栈

将其视为全球去中心化的开放计算基底。
“Raw atoms”
What it is:
  • GPUs, CPUs, memory, bandwidth
  • Heterogeneous hardware (consumer → data‑center grade)
  • Power, latency, geographic placement
Livepeer Context:
  • Hardware is contributed and owned by independent operators
  • Capacity is surfaced to the network via orchestrator infrastructure
Not Livepeer‑specific:
  • NVIDIA, AMD, bare metal, cloud VMs
Who provides it:
  • Orchestrators (GPU operators)
  • Gateways (in some cases - depending on deployment)
Livepeer does not own infrastructure. It indexes and incentivizes it.
“Bits move”
What it is:
  • Video transport, chunking, codecs, streaming primitives
  • Frame-level + segment-level delivery
  • Peer discovery, routing, and delivery paths
Livepeer context:
  • Segment-based processing and delivery patterns
  • Low-latency media delivery is a core competence (“video DNA”)
  • Gateway-mediated delivery patterns (client/server hybrid)
Who provides it:
  • Gateways (as the primary network-facing media/API surface)
  • Orchestrators (as the execution endpoints)
  • Underlying Internet transport (TCP/UDP/HTTP/WebRTC etc.)
Key Idea:
This is where Livepeer differs from generic GPU clouds: it is built around media-native transport patterns.
“Who Actually does the work?”
What it is:
  • Distributed job routing + load balancing
  • Service discovery and selection
  • Capacity-aware and performance-aware execution selection
Livepeer context:
  • Orchestrator selection and routing mechanisms
  • Segment/task model: work can be broken into independently computable units
  • Execution endpoints expose pricing/fees and performance characteristics
Who provides it:
  • Orchestrators (provide execution + advertise capability)
  • Gateways / clients (initiate and route work via protocol/network rules)
Key Idea:
The network decides who executes; the protocol incentivizes honest participation.
“Coordination, Trust, Provenance”
What it is:
  • On‑chain logic that secures and coordinates the network
  • Cryptoeconomic primitives that align incentives
  • Work/role coordination, reputation signals, and network rules enforcement
Livepeer Context:
  • On-chain protocol + off-chain coordination patterns that:
    • allow permissionless participation
    • coordinate stake-weighted influence and selection
    • align behavior with network goals
Who provides it:
  • Ethereum smart contracts (on Arbitrum L2)
  • Blockchain network (Arbitrum L2)
  • Network participants (orchestrators, delegators, gateways) operating under protocol rules
Key Idea:
This is the “brain” that replaces a centralized cloud control plane. It incentivizes correct behavior, but does not run jobs.
“Why anyone behaves” (incentives alignment, game theory)
What it is:
  • Cryptoeconomic primitives that align incentives
  • Game‑theoretic mechanisms that ensure correct behavior
  • Incentives for participation and penalties for misbehavior
Livepeer Context:
  • LPT staking + delegation
  • Inflation-based rewards (protocol issuance)
  • Fee capture and distribution (usage-based)
  • Slashing (where enabled) to penalize malicious/incorrect behavior
  • Governance processes (LIPs, voting, treasury stewardship)
Who provides it:
  • Protocol participants:
    • Delegators (stake delegation / security backing)
    • Orchestrators (service provision under stake)
    • Governance participants (proposal + voting + stewardship)
Key Idea:
Security is achieved via economic incentives + cryptographic verification, not trust in a single operator.
“What persists”
What it is:
  • The data plane around jobs: segments, manifests, logs, receipts
  • Temporary + derived artifacts needed for execution and verification
Livepeer Context:
  • Livepeer is intentionally light on storage
  • Video segments and state are often ephemeral and pipeline-driven
  • Long-term storage is typically external (object stores, CDNs, decentralized storage)
What persists:
  • Video segments (temporary)
  • Manifests, logs, receipts
  • Verification metadata
Who provides it:
  • Broadcasters / application operators (sometimes own long-term storage)
  • External storage providers (S3, GCS, IPFS, Arweave, etc.)
  • Not the Livepeer chain
Key Idea:
Livepeer coordinates compute over data, not long-term data custody.

​
平台与应用栈

“Batteries included”
What it is:
  • Managed services that abstract protocol/network complexity and make it easy to build on Livepeer
  • Operational tooling: dashboards, analytics, key management, billing abstraction
  • Provides features such as stream management, analytics, key management, billing abstraction
Livepeer Context:
  • Hosted/managed gateway experiences
  • Templates, pipelines, and reference apps
  • Ecosystem-led products that package the network into a usable platform
Who provides it:
  • Livepeer ecosystem teams (Inc, Foundation-funded projects, community teams)
  • Gateway operators offering service layers on top
Key Idea:
This is where the network meets real‑world needs. Products are opinions about access—they are not the protocol and not the network.
Examples:
  • Livepeer Studio
  • Streamplace
  • Daydream
  • and anyone who builds on top of the Livepeer protocol and network
See More Products & Platforms
“How builders integrate”
What it is:
  • APIs, SDKs, client libraries, docs, example repos
Livepeer Context:
  • Gateway APIs
  • SDKs (JS, Python, etc.)
  • BYOC-style integrations (containerized apps connecting to the stack)
Who provides it:
  • Product teams and open-source contributors
  • Gateway implementations exposing consistent interfaces
Key Idea:
This layer collapses network/protocol complexity into stable developer primitives.
See the Showcase
“What builders ship”
What it is:
  • End-user applications and services built on top of Livepeer
  • The final layer is where the end-user applications and services are built.
  • This is where the magic happens and is what most users will interact with.
Who provides it:
  • Independent developers, studios, ecosystem teams
  • Anyone who builds end-user applications on top of Livepeer or Livepeer Platform services
Examples: Decentralized Twitch alternatives, AI video apps, surveillance/vision pipelines, Metaverse/XR video. */}
Key Idea:
This is where the end-user applications and services are built. Applications are plural and replaceable; the network remains the substrate.
See the Showcase
“Humans and (increasingly) AI Agents”
What it is:
  • Viewers, creators, AI consumers.
Livepeer Context:
  • Users typically never see
    • orchestrators,
    • staking/delegation,
    • protocol mechanics.
Who provides it:
  • The applications and experiences (Layer 9)
Key Idea:
End users are the ultimate beneficiaries of the network, but they don’t need to know or care about the underlying technology. Success means protocol mechanics disappear behind great UX.
End users never see orchestrators, stakes, or protocol mechanics.

​
中间件与集成

Livepeer replaces persistent storage coupling with streamed, ephemeral compute coordination.
“joining layer”是： 协议层编排、基于分段的流式传输、经济保障。
Not Kafka. Not Airflow. Not Kubernetes. But conceptually: A decentralized serverless GPU fabric with a cryptoeconomic control plane.
Last modified on March 1, 2026