Episode 10 - Sept 2026 AI Roundup

FIELD NOTES // Episode 10 SEPT 2026 AI NEWS ROUNDUP: Progress, Risk & Who Do You Trust? DATE: September 2026 LOCATION: Somewhere between the Jetsons and the Matrix OBSERVATIONS • AI is moving extraordinarily fast—and frontier models are publicly raising questions about whether development should slow down. • Even if slowing down is desirable, actually doing it may be much harder. There are business and geopolitcal incentives to keep pushing forward. • AI is increasingly helping build AI. Coding agents can now write substantial amounts of software, assist with research, review their own work, and dramatically increase what small teams can build. • The same capabilities that create enormous productivity gains raise the question: at what point does the human in the loop stop understanding what they’re approving? • As AI assistants gain access to more of our lives, trust may become as important as capability—especially when personal data, financial information, advertising, and recommendations start converging. DISCUSSION In the first edition of a new recurring segment, Brandon, Jarad, and Travis tackle some of the biggest stories and questions in AI. First: should we deliberately slow AI development? The guys debate whether that’s even possible in a competitive global environment, the risks of increasingly autonomous AI agents, and whether incentives complicate warnings coming from the companies building the most powerful models. Next comes recursive self-improvement and AI-assisted development. If AI can help build better AI, and engineers become reviewers rather than writers, does that unlock an unprecedented productivity boom, or eventually leave humans approving systems they no longer fully understand? Finally, they dig into AI, advertising, privacy, and trust. Would you give an AI assistant access to your email? Your finances? Your purchasing history? And if its recommendations are influenced by advertising dollars, can you still trust the answer? As Travis puts it, trust may build gradually as these systems prove themselves—but discovering hidden financial incentives behind recommendations could destroy that trust just as quickly. KEY TAKEAWAYS → The AI race has a coordination problem: slowing down only works if everyone has an incentive to participate. → AI’s upside and downside come from the same thing—rapidly expanding capability. → AI-native workers may operate very differently from today’s workforce, with humans increasingly directing and reviewing work rather than creating every artifact themselves. → Human oversight only works when the human still understands enough to meaningfully oversee the machine. → Trust could become one of the defining competitive advantages in AI. The more access an assistant has to your life, the more important its incentives become. → And yes, apparently the Jetsons and the Simpsons may have predicted our AI future better than expected. LIGHTNING ROUND The guys close with AI clones, AI parenting, AI tutors, entrepreneurial children, smart refrigerators with judgmental personalities—and the important question of which responsibilities we’d happily outsource to machines. FOR FURTHER STUDY As AI becomes more capable, ask yourself: Where do you want AI to make decisions for you, where do you want it merely advising you, and where do you still want a human firmly in control? WHERE TO FIND US Brandon - https://www.linkedin.com/in/brmiddle/ Travis - https://www.linkedin.com/in/tlzimbelman/ Jarad - https://www.linkedin.com/in/jaradbacklund/ Jarad’s Writing - www.myoldnotebooks.com END OF NOTES