In this episode I sat down with my friend Cody Schneider and we live-built a GTM engineering “marketing machine” using Claude Code, MCPs, and a stack of APIs. The point isn’t “vibe code anything.” It’s: how do you get customers 24/7 by offloading the middle work (everything that used to require touching a keyboard) to agents. We spun up a bunch of Claude Code instances at once and started chaining real workflows together—LinkedIn auto-responders for asset requests, bulk Facebook ad creative generation, cold email automation, and even Notion doc creation—while other agents kept running in the background. The big takeaway: domain expertise + AI is the advantage. You don’t need to give an AI god-mode access to your life. You build specific jobs-to-be-done workflows, wire them to robust APIs, and then deploy them so they run in perpetuity. We talked through the foundational setup (a folder/repo you “live out of,” an environment file with API keys, and voice transcription like Super Whisper), then showed how you can generate and upload ad variants, track results with dashboards, and start moving toward a loop where agents create, publish, analyze, and prune winners/losers. This is what GTM engineering is becoming: personal software that compounds.