In his latest blog post, titled “Reflections”, Sam Altman, the CEO of OpenAI, wrote that they are now confident about their ability to build AGI (Artificial General Intelligence). That’s a pretty serious statement to make, and I don’t think Altman would make that jokingly. However, something else captured my attention. In the same blog post, Altman also said that OpenAI believes that in 2025 we will probably see the first AI agents join the workforce. That’s huge, especially because it is timed to happen in 2025 and not some distant future date. It prompted me to resume my experimentation with AI agents. In this article I am going to share what I think and what I learned. I will present my experimentations using DSPy, LiteLLM, Nebius AI Studio, OpenAI’s Swarm, CrewAI, and LangGraph.