In this Opto Sessions episode, I sit down with Amir, the co-founder of Humblytics—an AI-native analytics and conversion optimization platform—and a repeat founder with two exits. The goal of this conversation is simple: cut through the “AI hype vs. reality” noise and show what AI-native actually looks like in practice. We walk through the evolution from using LLMs as a helper, to automation, to augmentation/agency, and ultimately toward more autonomous agents—using the framework Amir credits to Anthropic. The most useful part is the concrete workflow: Amir runs his operating system in Obsidian (markdown files he owns), then layers Claude Code on top to turn that knowledge base into execution—outbound campaigns, lead enrichment, and even programmatic Meta/Google campaign setup via APIs and MCP/tooling. We also get into scraping and competitive research with tools like Firecrawl/Appify, and why “human-in-the-loop” is still non-negotiable: you’re accountable for the output, and strategy + taste still matter. The real limiter isn’t model capability—it’s culture, enablement, and leadership removing red tape so teams can actually learn, sandbox, and ship.