Telegram Cocoon is now online; your graphics card can now mine TON!

Yesterday, Telegram founder Pavel Durov announced on Twitter that their decentralized privacy computing network, Cocoon, has officially launched.

The Cocoon network consists of three parts:

Client: Sends a work request to the agent and pays the agent a fee after the request is completed.

Proxy: Routes work requests to worker nodes. Protected by a TEE, the proxy selects suitable nodes based on the worker node's device model, load, and reputation, and pays the fees collected from clients to the nodes. The nodes also take a commission. Telegram plans to allow anyone to run their own proxy in the future, making the Cocoon network completely decentralized.

Worker node: Executes AI inference requests in a TEE-protected virtual machine, and the agent pays the request fee from the client after execution.

In other words, anyone can earn $TON by using their own GPU as a worker node. Cocoon allows GPU owners to contribute AI computing power to the decentralized network on TON. By simply running the Cocoon protocol stack on a GPU server that supports TEE, they can provide private, verifiable AI model execution and transparently receive $TON payments for each processed request.

Pavel Durov stated that some AI-related features of Telegram, such as message translation, are already partially supported by Cocoon. He also mentioned Telegram's features such as speech-to-text and summarization, indicating that Cocoon did not lack practical applications from the outset, and believes that Telegram mini-programs will bring more usage demand to the Cocoon network in the future.

Compared to Telegram gift NFTs that previously generated hype in the cryptocurrency world, Cocoon is more like a decentralized AI computing network built by Telegram, combining blockchain and $TON as a payment method, starting from the built-in functionalities of Telegram. It's unlikely to issue its own token, limiting its potential for speculation. However, this project may reflect a long-term trend: increasing concern about privacy when directly exposing data to centralized AI companies during AI usage. In the cryptocurrency space, we've already seen more and more privacy-focused projects touting "protecting data sovereignty." (Related reading: 12 New Projects, a List of Privacy Projects Officially Mentioned by Solana )

Currently, the Cocoon network has a total TVL of approximately 4487 TON, with 30 worker nodes, 18 agents, and 12 customers. This data is from just over a day after its official launch, and it is still in its very early stages.

Supported by the practical application needs of Telegram's built-in functions and the TON incentive, the existence of the TON incentive also reduces the expectation of issuing tokens. Whether the Cocoon network can develop at a different pace from other crypto computing projects in the cryptocurrency circle still needs to be tested over time.