I tried to get a PhD in AI. Two times actually. First time was in 2001 right after I graduated. It didn't happened – work got in the way. Second time, in 2017, I was seriously considering it again. A PhD student at TUM Munich told me something that I still remember: "Eduard, is better to build real AI systems than to write papers about them." I took that advice. Then in 2022 I was preparing for my 3rd CCIE lab attempt. And I stopped. I realised I was optimising for a world that is already changing – fast. So I chose AI instead of the lab. All in. 25 years of networks. Started with missile and radar systems and NATO communications in Romanian military. Then 10 years at AT&T designing infrastructure for European banks and telcos. Kyndryl after that. Hundreds of enterprise environments across Europe. In 2018 I stood on a stage in Berlin and told a room of network engineers that ML will transform how we run critical infrastructure. In 2019 in Munich I showed how a trained neural network can predict and prevent failures in IP networks before they happen. Most of them didn't believe me then. Last year I built a 5-agent autonomous NOC – locally fine-tuned LLMs, no cloud dependency – that detects and resolves network incidents in seconds. Then a SOC system. Then a digital twin with 3 layers of validation before any change touches production. Then an AI security and compliance scanner that runs MITRE and NIST checks automatically. What human engineers were doing in days, the systems do it in minutes. No PhD. No research lab. Just 25 years of real networks and an obsession that never went away. Now I am building a community for MSP owners and IT consultancies that feel the same – that the ground is shifting and the old ways will not scale anymore. Is called Autonomous MSP. Is free to join. Link in bio.

A 3-person MSP. Friday. 11pm. 847 alerts in the queue. 3 active incidents. 1 engineer on call. A bank client on the phone. This is not a horror story. This is Tuesday for most MSPs I talk to. And here is what nobody says out loud – the engineer is not slow, is not bad, is not "not enough". The environment is just growing faster than any team can hire. In the last 5 years the average MSP environment complexity increased 3x. The alert volume increased 5x. The headcount? Maybe 1.3x if they were lucky. You cannot hire your way out of this. The math doesn't work. Alert fatigue is not a people problem. Is an architecture problem. The MSPs that will survive the next 5 years are not the ones with more engineers. Is the ones with better systems. AI-powered NOC that triages automatically. Agents that resolve before the human even sees the alert. Compliance checks that run without someone spending 3 days in spreadsheets. I am building exactly this – and I document everything live inside Autonomous MSP community. Free to join. Link in bio.

Nobody talks about what happens to the engineer after the NOC automates itself. I will tell you what I think. You don't get replaced. You get promoted. From managing routers to managing the agents that manage the routers. Think about it like this – 20 years ago the network engineer was manually configuring every interface. Then came templates, then automation, then intent-based networking. Each time, the engineer didn't disappear. They moved up one layer. AI is just the next layer. The engineer who understands how the AI agent makes decisions, when to trust it and when to override it, how to train it on your specific environment – that person becomes the most valuable in the room. Not because they write Python. Because they understand what the agents should and should not do. I call this the "Manager of Agents" role. And I think every serious network and security engineer needs to start thinking about this transition in the next 2-3 years. Is not optional. Is the same as when routing protocols replaced static routes. You adapt or you fall behind. The good news – you already have the hardest part. 10, 15, 20 years of real network experience. That knowledge is exactly what makes a great agent manager. The AI doesn't know your environment. You do. I built Autonomous MSP community specifically for engineers and MSP owners making this transition. Is free to join. Link in bio.

DRAFT — To be written after Cisco DevNet podcast. Content brief: Live demo of the 5-agent NOC. Reference the podcast appearance. Show real metrics — detection to resolution time. End with community link and pilot CTA.