Photo by energepic.com on Pexels

This week witnessed a flurry of activity in the Agentic AI space, marked by groundbreaking advancements and emerging conflicts. A dedicated search engine for AI agents has launched, promising to streamline agent workflows. Meanwhile, the Chinese AI model, K2 Thinking, has reportedly outperformed GPT-5 on the Humanity’s Last Exam, achieving a score of 51%. Further fueling the competitive landscape, Amazon is suing Perplexity over its AI shopping agent, raising questions about the future of AI regulation. Microsoft Research unveiled its “Magentic Marketplace” simulation, highlighting potential weaknesses in agentic markets. Perplexity also released significant upgrades to Comet, enabling it to handle more complex workflows. The week’s developments also include Google AI’s introduction of an Agent Development Kit for Go and the release of new tools for testing and scaling AI agents. Salesforce boosted its enterprise AI agent capabilities with the acquisition of Spindle AI, now integrated into Agentforce. As Black Friday approaches, Microsoft is enhancing Copilot Shopping with new features. Lastly, Runable introduced an agent capable of handling diverse media formats. This rapid evolution underscores the transformative potential and growing complexities of Agentic AI.