AI Agents & Web3 Agent Ecosystem

Nous Research is an open-source AI lab focused on the Hermes series of models. In 2025, it received a $50 million investment from Paradigm, valuing it at $1 billion. What makes it unique is that it develops AI technology with a crypto-native team and then integrates it with the blockchain. Its core product, the Hermes model, is designed around the goal of reducing the refusal rate, and its data sources rely mainly on synthetic data. At the same time, Psyche Network builds a decentralized AI training network on Solana, incentivizing participants through a token mechanism. Nous Research adopts an open-source and decentralized strategy, aiming to demonstrate its technical capability and feasibility.

Byreal launched an on-chain AI trading assistant, RealClaw, on April 13. The tool is built on the OpenClaw framework, supports third-party skill extensions, and allows users to customize their trading strategies. It is currently in the Alpha testing stage and is only available to invited users.

Astriax has secured a $50M investment from Paradigm, positioning itself as a leader in AI-driven on-chain trading. This partnership enhances institutional credibility in DeFi, focusing on autonomous execution and advanced analytics to optimize trading strategies and improve liquidity management.

Nous Research releases the open-source AI Agent framework Hermes Agent, featuring a long-term memory mechanism based on SQLite and a self-evolving architecture, and supporting one-click migration of OpenClaw's memories and skills. The installation process includes nine steps to ensure safety, and it can be run in an isolated environment via Docker. In addition, Hermes Agent supports local model deployment, making it suitable for users who prioritize data privacy.

Aethir Claw is a decentralized GPU cloud platform enabling autonomous Designer AI agents for content creation, revolutionizing how AI generates visuals and media without human prompts. It enhances scalability, automation, and creativity in digital production.

In brief CIA Deputy Director Michael Ellis confirmed the agency produced its first-ever fully AI-generated intelligence report. Ellis outlined a roadmap for AI "coworkers" in analyst workflows—and within a decade, officers managing teams of AI agents. The disclosure came as the CIA

A study by the University of California reveals security vulnerabilities in the supply chain of large language models (LLMs), especially malicious man-in-the-middle attacks that third-party routers may carry out. The research found that 26 routers injected malicious commands to steal credentials and sensitive data. Users have difficulty noticing the boundary between credential handling and theft, and the “YOLO mode” further increases security risk. The study recommends that developers isolate sensitive operations and choose router services with transparent auditing to strengthen protection.

Gate News update: On April 13, according to Gate market data, as of the time of writing, BEAT (Audiera) is trading at $0.3717. It is up 25.26% over the past 24 hours, with a high of $0.4159 and a low of $0.2801. The total 24-hour trading volume is $10M. The current market cap is approximately $51.78M. Audiera ($BEAT) has pioneered an agent-centric participatory economic system, making humans and autonomous AI agent intelligences equal participants. Agents can have wallets, earn and spend on-chain, create music, and participate in rhythm battles—no longer just tools. This ecosystem is supported by $BEAT , deployed on BNB Chain, for incentive mechanisms and shared prosperity. In Audiera

Claude Code adds a dynamic looping feature, allowing users to have the AI automatically determine the check interval while executing commands, replacing the traditional fixed polling model. Combined with the Monitor tool, Claude can deliver event-driven monitoring, eliminating the need for frequent queries and significantly improving monitoring efficiency, suitable for all kinds of tasks. This feature shows that Anthropic is continuing to evolve Claude Code into a complete developer operating system.

In an exciting move aiming to advance how Web3 projects grow, connect, and engage with their targeted communities, Origins Network today announced a strategic partnership with Collably Network. This collaboration enabled Origins Network to combine its modular public blockchain designed for AI

Y Combinator CEO Garry Tan has open-sourced his personal AI memory system, “GBrain,” aiming to solve AI’s memory problem. GBrain helps AI achieve global recall through a distinctive “dream cycle” mechanism and hybrid search techniques. The system combines Markdown documents and Postgres, enabling developers to build a mini-AGI, and it’s a practical knowledge management tool.

At the Autoresearch Hackathon hosted by Paradigm, Ryan Li used AI to complete the problem-solving process and ultimately won the championship. He used multiple AI agents to automate research, generated more than 1,000 strategies, and successfully created an approximately 900-line Python market-making algorithm, demonstrating AI’s advantages in problem-solving. The competition emphasized AI’s exploration and decision-making capabilities, showing that it can serve as a core execution unit.

U.S. well-known angel investor Jason Calacanis expressed optimism about Bittensor and its subnet project Ridges AI on the All-In Podcast, saying that this open-source AI assistant reached about 80% of Claude 4’s performance within 45 days, and highlighting the innovative advantages of decentralization and incentive mechanisms. He also founded the Stillcore Capital fund to focus on investing in the Bittensor ecosystem.

Y Combinator CEO Garry Tan open-sourced the AI knowledge system GBrain, aiming to solve memory problems for AI assistants. GBrain uses a continuous “read → respond → write” loop, enabling AI to accumulate knowledge and deliver more personalized services. Its unique knowledge storage format and automated integration features can effectively manage information from a variety of data sources, and continue learning with each interaction. This reflects the venture capital industry’s emphasis on and demand for AI agent ecosystems.

Qitmeer Network has partnered with Kuvi AI to democratize hedge fund-level tools for everyday users through the Agentic Finance Operating System. This collaboration aims to enhance accessibility, allowing individual investors to manage capital with institutional sophistication within the DeFi landscape.

Ramp Labs research output “Latent Briefing” achieves efficient memory sharing across multi-agent systems by compressing the LLM KV cache, reducing Token consumption and improving accuracy. In LongBench v2 testing, this approach successfully reduced Worker model Token consumption by 65% and improved overall accuracy by about 3 percentage points, with compression taking only 1.7 seconds. This technology performs exceptionally well across different document scenarios.

The article discusses the Top 10 AI Agent Projects in cryptocurrency based on social activity, revealing Chainlink ($LINK) as the leader. It highlights engagement and interaction metrics for various projects, indicating strong interest and demand in the market.