Hey @Roman Bediner - perfect timing. I've been planning to start building the “Agentic Content Engine” for our LinkedIn content agency. There is a HIGH bar here because right now, all posts have human writers & editors carefully reviewing content before it gets published to ensure tone of voice, natural speech patterns, etc. are carried through from raw call transcripts to final post. That's our real differentiator: the content need to remain unfiltered and undiluted. As potent as it was when it left the founder's mouth, just sharpened and structured into something people can actually consume on LinkedIn. The first step is still human: hours of conversation, intentional questioning, follow-ups, and digging to extract the founder's judgment, stories, worldview, natural cadence, and the things they know but wouldn't think to write down themselves. What I want to build is the proprietary agentic workflow that comes next. I'm imagining something like a series of specialized agents working against a persistent "Founder Brain": Raw interview transcripts → insight/angle extraction → deeper interrogation where context is missing → draft using primarily the founder's own language → editorial council for voice, narrative, clarity and potency → recursive revision → human approval. V2 would have a learning loop too... Every time I or an editor changes the machine's output, the system should compare the original draft against the approved version, abstract why those changes were made, and -with the right guardrails - update what it knows about that founder and our editorial standards. So instead of giving the same feedback over and over, each approved post makes the system better at writing the next one. Over time, we'd effectively be building three layers of intelligence: 1. The founder's judgment, stories, worldview and voice. 2. Frontier's editorial judgment around what makes an idea sharp, credible and worth publishing. 3. Eventually, performance feedback around which posts resonate most with the right buyers.