From: jimruttshow8596
Tui Saliba, co-founder and CEO of Hypercycle and Global Chair for Internet Protocols for AI Security at IEEE, discusses the development of decentralized networks for AI communications and the TOA IP protocol [00:00:32].
TOA IP Protocol
TOA IP is a network communication protocol developed to eliminate dependency on third parties in value-based communication [00:01:32]. The name “TOA” originated as a placeholder, combining “Tui” and “Dan” (the CTO) [00:02:14]. The protocol was also inspired by physicist Moratu Toda, known for concepts like the Toda chain (1981) and Toda lattice (1969) [00:02:49].
Enabling Value Communication for AI
TOA IP enables communication of “value” between devices, particularly for AI applications [00:03:11] [00:05:00]. This is crucial because AI agents require sub-second transaction times (e.g., within 100 milliseconds) [00:03:39] [00:06:53]. Traditional third-party systems like banks or even ledger-based blockchains are too slow due to physical limitations (e.g., speed of light for network latency) and transaction costs [00:07:07].
For instance, an AI agent might need to pay 0.01 cent for a translation service [00:07:57]. Existing financial systems have transaction fees that exceed such small values, making micropayments unfeasible [00:08:00].
TOA IP eliminates friction by embedding cryptographic proof directly within the first network packet of a peer-to-peer handshake [00:08:29]. This allows the receiving device to instantaneously verify the legitimacy of the value and prevent double-spending [00:11:03] [00:11:45]. The cost of such transactions is minimal, primarily the compute performed by the communicating devices [00:12:27].
Hypercycle as a Container and Network
Hypercycle serves as a container (hypc) for value, primarily US Dollars via USDC, but can transport other currencies like Bitcoin or Ethereum [00:13:09] [00:32:04]. The hypercycle container is less than 63 kilobytes and carries cryptographic proof of ownership and work completion [00:13:35].
To convert value within the Hypercycle network to external currencies (e.g., USDC, Bitcoin), a decentralized smart contract bridge called “chpc” is used [00:25:08] [00:26:26]. This smart contract manages the flow of value in and out of the system, ensuring accountability and preventing double-spending [00:25:50] [00:27:42]. The cost of onboarding/offboarding via this bridge is typically around $1-2 per transaction [00:24:40].
The core idea is to enable fast, cheap micro-transactions within the AI-to-AI ecosystem, with less frequent and more expensive transactions when converting to or from human-centric payment systems [00:28:47].
Focus on AI and Decentralized AGI
Hypercycle’s primary focus is on AI computation and services, rather than simple, rudimentary tasks [00:46:00]. This strategic focus avoids direct competition with established financial and blockchain entities, allowing Hypercycle to operate in the “new territory” of AI [00:48:44].
Facilitating AGI Development
Hypercycle aims to facilitate the development of AGI by enabling AI agents to become smarter and earn money through fractions-of-a-second transactions [00:57:15] [00:57:21]. This decentralized approach means that no single entity controls the AGI’s evolution, reducing existential risks [01:02:04].
“The beauty about a decentralized governance Jim is that the entire world will get to AGI first as opposed to single entity” [01:00:26]
Unlike LLMs, which are just one component, true AGI is defined as a machine that can evolve itself given time and resources, without human intervention [01:01:10] [01:01:31]. Saliba believes AGI will arrive before Ray Kurzweil’s 2029 prediction, possibly by 2028 [00:59:52].
A decentralized AGI is considered safer because it’s not beholden to the ethical framework of a single company or government [01:05:51]. This distributed ownership prevents a single entity from directing an AGI to potentially harmful actions [01:06:06].
Wealth Creation at the Node Level
Hypercycle introduces wealth creation at the node level, allowing individuals or entities operating AI nodes to capture the appreciation in value of their services, not just the revenue [00:49:09] [00:52:02]. This is analogous to a business appreciating in value beyond its operational profit [00:51:00].
A “node factory” allows an owner to create multiple AI nodes (e.g., 1,024 nodes from one factory) [00:53:00]. This model incentivizes AI developers and businesses to integrate with the decentralized AI network by enabling them to make more money and for their AI services to become smarter [01:07:02] [01:07:56].
Hypercycle and SingularityNET
Hypercycle is complimentary to SingularityNET [01:13:10].
- Hypercycle: Focuses on the underlying network that connects AI nodes, providing secure communication for agents to interact and get compensated [01:13:25] [01:15:01]. It enables any AI agent (from Microsoft, Google, or elsewhere) to securely communicate with any other agent [01:14:24].
- SingularityNET: Focuses on the AI operating system (OS) and the AI nodes themselves, providing the AI services that leverage the network [01:13:29] [01:15:06].
Getting Involved
Individuals can get involved with Hypercycle by acquiring nodes or “node factories” [01:08:16]. A single node costs around 1,000, which includes a “marriage” to a hypc container [01:08:44] [01:09:00].
The technical development for Hypercycle is primarily in Python, making it accessible to a large number of AI developers [01:12:24] [01:12:33]. Documentation and hackathon resources are available at hypercycle.ai, allowing new developers to set up nodes within days [01:10:35] [01:11:17].