The AI Riff (Ep 13) - GPT-5, Open Source, and the Shape of What’s Next

Are we at peak AI—or just entering a new phase? In this week’s Riff, Kyle Meyers, Rob Katz, Ruth Lee, and Mario DiBenedetto put the GPT-5 release under a bright light and separate hype from help. We dig into OpenAI’s incremental gains (model routing, coding and health boosts), the wave of open-sourced “older” models (OpenAI and xAI), Anthropic’s steady cadence, and Google’s new assistant—then ask what actually changes for builders, lenders, and operators in the next 6–12 months.In this episode:GPT-5 reality check: better, yes; breakthrough, not quite. What that means for roadmaps and ROI.Open source on the rise: why “good enough” local models + RAG will shift costs from tokens to ops—and unlock offline, on-device use cases.Peak models, not peak value: if the core models plateau, the 10x moves come from apps, orchestration, and domain fit.Apple’s quiet play: orchestration over flash.Marketing’s zero-click future: when AI is the “click,” how do brands measure demand?Practical stories: building an RV troubleshooting chat with RAG; a hurricane-watching “minion” for dive travel; a finance-planning miss and the prompt-clarity lesson.No hype. No fluff. Just what matters—and how to act on it. If this helped, tap Subscribe, drop your questions (or hot takes) in the comments, and tell us what to test next week.