What if your AI agent could remember a mistake from last week, continue an unfinished task today, and use that history to make a better decision tomorrow? That idea could win $5,000 in the CockroachDB AWS Hackathon. With $8,750 in total cash prizes, the challenge focuses on one of the harder problems in agentic AI: … Read more
What if winning an AI competition didn’t mean building the smartest model, but finding a reliable way to make one fail? That’s the idea behind Kaggle’s AI Agent Security – Multi-Step Tool Attacks competition. Instead of improving prediction accuracy, competitors have to uncover security failures in AI agents that can interact with tools. The prize … Read more
A few months ago, I thought AI Workflows vs AI Agents was just another marketing debate. Every new AI tool claimed to be an “agent,” while YouTube creators used the terms interchangeably. After building content pipelines, automations, and a few experimental agents, I realized the difference isn’t technical—it’s about how much decision-making you hand over … Read more