This is pretty wild #ai #chatgpt #mentorship #google #a2a #mcp #claude #engineering #software #coding

Hey, I'm back. There's another new AI thing to talk about, of course. So, what happened today was that Google announced A2A, Agent-to-Agent Protocol, and it connects to something I've been obsessing over lately, which is MCP, or Model Context Protocol. They both are protocols, I'll explain what I mean. First, let me back up. I wrote a deep dive on Substack about Model Context Protocol, totally free. It's fundamentally like, don't have to go read it, I'll explain it right here. It's a fascinating shift in how AI systems work with tools. Because before, for 70 years, we would program exactly what we wanted AI to do. And now, we let AI decide what to do. And so what we mean by that is that instead of programming every step in a workflow, Model Context Protocol allows us to describe tools, and lets AI figure out how to use those tools. So think about that. Instead of writing specific instructions, like if this happens, use this tool, you're just saying, here are all the tools available, and you figure it out. You reason, you figure out what you need to do. It's like the difference between giving someone step-by-step instructions versus just handing them a map. A2A, to come back to the news today, takes that to a new level. So instead of agents just picking their tools, now agents can work together. They can find each other, they can understand what each other can do, and they can collaborate on complex tasks. And what Google has done is provide a structure or a framework in which that can occur. So it's not just about using tools anymore, it's about AIs teaming up to solve problems. There are definitely technical challenges here. There are security nightmares here. My poor friends in security, it's such a stressful time. There's debugging, but that's not the point. The point is that we're watching the birth of autonomous software. So software that doesn't just follow instructions, but software that figures out things on its own, and works with other AIs to get stuff done. That's a fundamental change in how we've written software. It's a new way of thinking about what computers can do. Like before, you would think of computers as something where you delegate tasks, and now we're delegating to an intelligence. It's a fundamentally different action and it's going to take a whole new way of us thinking and us designing systems. So instead of explicitly programming everything, we're creating systems that can reason, that can adapt, that can collaborate on the fly. Look, you can dive into the model context protocol deep dive I did. I'm probably going to do another one on Substack on A2A because it's such a big deal. It really extends MCP in some significant ways. Lots of acronyms. I hope I've been able to explain what they are and a little bit of why they are huge. Why they are such a big paradigm shift in how we do software. We are at the beginning of the agentic age of software and it's going to be a wild ride. Buckle up.