74% of shoppers have used an AI agent, but only 14% trust one to actually place the order. In episode ten of Breaking the Build, Tim and James dig into agentic shopping: whether anyone will really let an AI spend their money, why Amazon Dash buttons and Alexa ordering never took off, and who's on the hook when your agent buys something you never wanted.

Topics covered: agentic commerce, AI shopping assistants, AI agents, ecommerce fraud and chargebacks, AI product recommendations, AEO (answer engine optimization), paid placements in AI answers, Instacart and grocery delivery, voice shopping and smart speakers, AI adoption by generation.

This casual tech podcast splits shopping into the predictable stuff you'd happily hand off and the fun stuff you wouldn't, then asks what banks, insurers, and retailers will have to change before an agent can check out for you. Also: a closet full of Costco chips, ten jars of peanut butter, "agent jail," a vending-machine AI that went full Gilded Age tycoon, and the official rule that agents may only shop for you once you can hurt their feelings.

Perfect for: ecommerce professionals, retailers and marketers, developers building with AI agents, tech enthusiasts, anyone who has ever closed the Rufus pop-up on Amazon.

Tech's favorite gospel is "launch fast, iterate, see what sticks" — but Tim and James want to know who's actually running the beta test. (It's you. You're the beta test. You're also paying for it.)

This episode wanders from the death of the "viable" in MVP through the Google and Microsoft product graveyards, the tragic murder of shuffle-less music apps, and the Zune's doomed quest to be cool. Along the way: why Netflix's eight-episodes-and-cancel model may have killed the sitcom entirely, how streamers accidentally reinvented weekly television and called it innovation, and the slow horror of realizing you don't actually own your thermostat, your smartwatch, or your recipes. Plus the crab theory of software evolution, a genuine pitch to bring back Clippy, and three would-you-rathers about corporate abandonment that spiral exactly as far off-topic as you'd hope.

The theme, as always: at the end of the day, the customer suffers.

Should you say "please" and "thank you" to your AI? Tim does. James absolutely does not. In this episode of Breaking the Build, Tim and James debate AI politeness, token optimization, caveman speak prompts, and whether being rude to your LLM shapes the agent you're building over time. Plus: the Gemini calendar incident, AI as therapist, and James's plan to outsource his mean prompting to a dedicated executioner bot.

Topics covered: AI politeness, token usage, prompt engineering, LLM behavior, caveman speak, anthropomorphizing AI, AI addiction, Claude Code, ChatGPT, Google Gemini, AI tools for developers.

Perfect for: developers and tech professionals who use AI tools daily, anyone curious about how prompt style affects AI output, and people who have ever felt vaguely guilty about being rude to a chatbot.