Dymension has basically launched the equivalent of https://t.co/f82OHgafMn, but for launching your own rollup in 1-click that can have its own apps (rather than just being a token), using Celestia as DA: https://t.co/CxbCajLex3 ๐
Awesome to see our new consensus protocol Autobahn having impact in industry! (@ this year's SOSP: https://t.co/gTryChs3FE)
Excited to see Somnia power @Improbableio's Msquared : https://t.co/B2AbeXbr2F
PS: Hire Florian, Neil, the brains behind Autobahn! @ittaia@floriansuri
Translated the groundbreaking paper introducing a trustless IP exchange protocol for AI agents, "Agent TCP/IP," by @devrelius and @jasonzhao of @StoryProtocol into Korean. This translation was drafted using ChatGPT 4o and reviewed/modified with figures.
The pdf/latex files can be found in https://t.co/AkzMS4mtsE.
์คํ ๋ฆฌ ํ๋กํ ์ฝ์ @devrelius@jasonzhao๊ฐ ์ด Agent TCP/IP ๋ ผ๋ฌธ์ ํ๊ตญ์ด๋ก ๋ฒ์ญํ์ต๋๋ค. "Agent TCP/IP"๋ AI ์์ด์ ํธ ๊ฐ ์ ๋ขฐ ์์ด IP ๊ตํ ํ๋กํ ์ฝ์ ์ ์ํ ํ์ ์ ์ธ ๋ ผ๋ฌธ์ธ๋ฐ, ๋ ๋ง์ด ๊ณต์ ํ๊ณ ์ถ์์ต๋๋ค. ์ฑGPT4o๋ก ์ด๋ฒํ๊ณ , ๊ทธ๋ฆผ๊ณผ ์ด์ํ ํํ ๋ฑ์ ์์ ํ์ต๋๋ค. ์ ๊นํ๋งํฌ์์ pdf/latex ํ์ผ ๋ฐ์๋ณด์ค ์ ์์ต๋๋ค.
Introducing Agent TCP/IP:
An experimental framework that lets agents trade IP with each other.
Agents can now exchange their training data, creative style, investment strategies, and more, ushering in an intelligence economy. ๐
Reviewing @Somnia_Network's consensus mechanism, I'm quite impressed. Its principle looks similar to DAG-based consensus like @Aptos and @SuiNetwork (@Mysten_Labs).
Somnia's MultiStream consensus (https://t.co/ojYZJ023N5), inspired by Autobahn BFT (https://t.co/JKyGxawpVg), gains performance by "every validator publishes their own blockchain." This approach is strikingly similar to Narwhal but evolves from the likes of Tendermint (CometBFT) and Hotstuff, making it straightforward and easy to understand.
While traditional consensus like Tendermint/CometBFT integrate consensus and execution at one binary, others have gained by separating these layers.
DAG-based and MultiStream go further, splitting the data (transaction) dissemination layer from the consensus layer for even greater performance benefits. As consensus mechanisms for blockchain continue to evolve in different directions, it feels like they've been converging towards a single point.
Really looking forward to Somnia's public testnet/mainnet launch!
if you could customize one thing in any of the existing L1/L2s what would you do?
what does it unlock to that L1/L2?
anything is fair game. new opcode, tx type, precompile, vm, centrized stablecoin issuance api etc
๐ฃ๏ธBig announcements
๐Major milestones
๐คGroundbreaking tech rollouts
& much more!
Gather round the fire, Sui fam, and warm yourselves up with our 10 hottest X posts of 2024! ๐ฅ๐
https://t.co/YQ4v0mTBv5
You don't need to remove "consensus" to optimize a DA layer. Consensus isn't the bottleneck, block propagation is. You can just separate consensus from block propagation. Which is what we do here: https://t.co/1QOqhk5WhX
Really promising research that could allow us to validity prove the Celestia erasure coded data, while keeping the existing fast Merkle tree based commitment scheme, instead of needing to switch to (slower) polynomial commitments!