Translate to Simplified Chinese Language: • %% • %% Because you’ve shown interest in a similar community %% I’m trying to automate an entire AI company. 60% done, getting exponentially harder. Is this even possible?” %% So I’m 17 and I’ve been building AI automation systems for a few months. Started with basic workflows, now I’m deep in something that might be insane. The goal: Automate a full AI infrastructure company. Not like “automate customer support” – I mean everything. What’s already running: %% Sales teams that research leads, scrape data, build personas, and handle my clients %% Content systems cranking out newsletters, social posts, podcast scripts at scale %% Multiple RAG systems with metadata and vector stores handling different knowledge domains %% currently 122 specialized agents with dedicated software functions or just regular n8n agents %% Multi-contextual agentic systems (agents that understand business context across departments) %% Several MCPs orchestrating agent-to-agent communication %% Every automation step generates training data for the next iteration %% n8n infrastructure team. that basicly does my job. and a fullstack developer for creating app around workflows. %% Basically: systems that build systems that train systems. The last 40% is breaking my brain. Once you have 122 agents that need to coordinate, everything gets exponentially complex: but i need to scale it to 280-300 agents or systems. %% Sales agent needs context from content systems %% Content systems need data from sales conversations %% Research agents need to feed multiple departments %% Decision-making needs to happen across disconnected workflows %% Agents stepping on each other / duplicate work %% Context windows maxing out %% State management across sessions %% Training loops creating feedback without human validation %% It’s not that any single piece is impossible. It’s that coordinating autonomous intelligence at scale is a fundamentally different problem than building individual automations. Like the difference between building a car and building traffic infrastructure for a city. My actual question: Do you think it’s possible to fully automate an AI infrastructure company? Not theoretically – I mean in practice, with current tech. Where’s the ceiling? Is 80% realistic? 95%? Or is there a fundamental limit where human decision-making becomes mandatory? Either I’m building something genuinely new or I’m about to learn why nobody else has done this. Figured Reddit would have opinions. %% Look, I get it. You’re shipping fast. But let Coderabbit review every pull request. It’ll catch SQL injection, exposed credentials, broken auth before anything goes live.