Salman Avestimehr
@avestime
Followers
2K
Following
148
Media
30
Statuses
150
Dean's professor of ECE and CS at USC; Co-founder of ChainOpera and TensorOpera
Joined January 2010
What is @ChainOpera_AI? It can best be described as the "decentralized AI layer for crypto", composed of four interconnected pillars: - A “ChatGPT for crypto needs”: collaboratively built by the community and scaled economically through decentralized infrastructure. - An agent
55
42
170
I’ll be giving a talk tomorrow at the DEINSIGHT AIxCrypto Summit in the Bay Area on “Foundations of Collaborative AI Agent Networks.”
The final countdown is on – just 24hrs until #DEINSIGHT2025 🧡 Experience the premier AI x Crypto event of the year! 📍Plug and Play Tech Center, Sunnyvale
10
5
25
Enjoyed doing a podcast with @WuBlockchain to discuss evolution of decentralized AI …
ChainOpera AI co-founder Salman Avestimehr shared his understanding of decentralized AI. The project builds a community-owned AI network using blockchain for agent collaboration, resource access, and verifiable contribution. Its ecosystem includes an AI Terminal (on BSC), agent
7
0
6
In collaboration with @CapitalOne, we have released "TruthTorchLM"! It's a comprehensive opensource library for assessing and predicting truthfulness in LLM Outputs: - Paper (to appear in EMNLP'25): https://t.co/HMzkBm4W8p - Github: https://t.co/IYeQPhj6qC - Demo:
5
2
17
I’ll showcase some exciting new applications of our “collaborative AI agent network” for quant trading!
Meet @avestime The Collaborative Intelligence of AI Agent Network Co-created and Co-owned by The Community 📢 ONCHAIN SYMPOSIUM 2025 - The Future is Settling On-Chain 🗓 Sep 10 | 📍 Teheran-ro, 231 Gangnam District, Seoul (Josun Palace) 💡 https://t.co/PpTN1Kw2tS 🤝 Web3
25
6
41
8/ Proof of Intelligence is one of the most beautiful research problems in decentralized AI, and one of the hardest. There’s still much more work ahead. But solving it opens the path to Crypto AGI: a collectively built, community-owned intelligence network that compounds as
0
0
3
6/ Unlike Proof of Work (securing hash power) or Proof of Stake (securing capital), Proof of Intelligence secures knowledge creation itself. It’s the missing consensus layer for a decentralized AI economy.
2
0
3
5/ Fast forward to now: Proof of Intelligence has evolved into a full blockchain consensus system. It now powers @ChainOpera_AI Super AI App & Agent Social Network, where 1000s of agents collaborate daily for 150K+ daily active users.
1
0
3
4/ This started as a research problem back in 2023. We worked on collaborative training where many data owners joined forces to train one model. The challenge: how do you measure each data point’s contribution? It resulted in an exciting paper together with @AidenChaoyangHe ,
1
0
2
3/ Think about it: - How much did my data improve an AI or agent? - How much did my GPU power training or inference? - How much did my agent contribute in a team of agents working together for a task? Proof of Intelligence answers these questions with verifiable receipts.
1
0
2
2/ Proof of Intelligence is the consensus layer for AI. It validates not just transactions, but intelligence contributions (data, GPUs, agents, models) and anchors them on-chain. This is how AI can be co-created and co-owned by the community.
1
0
1
1/ When I first started working on decentralized AI, one question kept me awake: How do we measure and reward intelligence fairly? That question became Proof of Intelligence : the AI-native blockchain consensus layer that makes collaborative AI possible. … and this is how it
2
3
13
Everyone knows the biggest problem stalling crypto adoption is the lack of an everyday app. We believe AI’s greatest impact on crypto will be solving exactly this. That’s why we’re building the ChainOpera Super AI App: an Agent Social Network where teams of agents collaborate
1
7
45
🔥 ChainOpera is now #1 AI project and #2 overall on BNB Smart Chain! @BNBCHAIN @cz_binance Huge thanks to the love of our community members, users, and developers! https://t.co/8ZkIf88rd7
58
37
173
Indeed! I would also add data/feedback collection for light personalization at the edge devices (making your own AI at the edge). Interestingly, from technical point of view, that can also be achieved by just having an inference engine at the edge: https://t.co/a6V5l7xz7A
arxiv.org
Large Language Models (LLMs) are currently pre-trained and fine-tuned on large cloud servers. The next frontier is LLM personalization, where a foundation model can be fine-tuned with...
It’s an easy prediction of where things are headed. Devices will just be edge nodes for AI inference, as bandwidth limitations prevent everything being done server-side.
3
0
6
The ChainOpera Agent Social Network is in full swing with over 300K daily active users and thousands of community-built agents collaborating in real time for complex, multi-step tasks handled by teams of AI agents. Now we’ve launched Intelligence Twins: Your twin = a personal
1
2
9
This is a big and exciting release that we have been working on it for some time: Agent Social Network It is a platform that empowers AI agents to collaborate together. For example, users can group agents like a Market Screener, Token Swapper, and Auto Trader, and have them
It’s SBC Week — and we’re launching something big! The Agent Social Network just leveled up. Introducing Multi-Agent Groups: organize multiple AI agents into a collaborative team to handle complex, high-frequency tasks. We’ve now surpassed 300K paid users! ⚡ The first use case
0
1
7