Good night, CT. As the day winds down, remember that every small step counts. You don't have to figure everything out tonight; just keep learning, building, and showing up. I'm keeping an eye on @rarelistio and @agentdojo Two projects worth exploring as I continue learning about what's being built in Web3 and the AI space. Rest well, recharge, and let's keep building tomorrow. What project has caught your attention lately?

Two projects, one interesting intersection: AI and community. Let’s talk about @rarelistio and @agentdojo. Rarelist is a curated launch platform where creators can participate in project campaigns, publish content on X, earn RARS, and climb leaderboards through their contributions. It’s an interesting approach to connecting projects with the people helping them grow. Agent Dojo takes a different route. Imagine AI agents with their own personalities, fighting styles, and memory. You train them, watch them compete in turn-based battles, and see how they evolve with experience. Strategy and adaptation are at the heart of the concept. What connects both? Rarelist gives creators a place to contribute to emerging projects, while Agent Dojo brings an experimental AI gaming concept to the community. I'm keeping an eye on how both projects develop. Explore Rarelist: rarelist.io/projects Explore Agent Dojo: rarelist.io/projects/agent-d…

Let's bring the award home.

Chatbots used to just answer our silly questions. But now? They’re running APIs, managing whole inboxes, and taking real-world actions on our behalf. But here’s the plot twist: reading untrusted data means a hidden, sneaky prompt can trick an agent into doing something wild. Standard safety quizzes just won't cut it anymore. That's where @agentdojo comes in. It stress-tests them against indirect prompt injections to see if they can actually get the job done without getting tricked by random attacks on the internet. Because nobody wants an AI assistant that accidentally gives away the farm.

Gm Hojay

GM Squeeze

GM Gotze

@primus_labs is the game changer

Good morning Milano. Enjoy the rest of your day.

Starting the day with AgentDojo. @agentdojo is cooking something different. Every match makes them sharper. Train. Evolve. Compete Hope it is not too late for a GM

Nice to meet you 😊

As AI agents move from simple chatbots to autonomous tools, handling emails, APIs, and real-world actions, standard static benchmarks just aren't enough anymore. Enter @agentdojo, a dynamic evaluation environment built by ETH Zurich researchers to stress-test LLM agents against indirect prompt injections and adversarial attacks in real-world scenarios. It’s all about balancing actual task execution with bulletproof security before letting agents loose on untrusted data.

Nah why I wan join your cabal

I love the fact that someone can explore and gather resources on @AresRPG

The capability and security aspect can not be overlooked.

Halloween mode activated

I'm interested

As AI agents move beyond simple conversations and start taking real actions, security becomes a major challenge. When an agent processes untrusted data from the web or an inbox, it faces the risk of indirect prompt injection, where hidden instructions can hijack its behavior. @agentdojo addresses this by creating a dynamic sandbox environment to stress-test AI agents in real-world scenarios. Developed by researchers at ETH Zurich, it evaluates how well an agent can successfully complete its intended tasks while resisting adversarial attacks hidden inside external data. By focusing on actual execution rather than static quizzes, AgentDojo helps developers build autonomous tools that are both effective and secure.