If you're building an AI product, interface is your primry competitive advantage (eleganthack.com)

Google’s new image generator (Nano Banana Pro) impressed in demos, but the author admits they instinctively switched back to Claude for a non-image task — not because it’s objectively better, but because they’re familiar with its “voice” and interaction model. This illustrates the 9x problem from John Gournville’s HBR paper: users overvalue what they already know ~3× and builders overvalue their product ~3×, so a new product must be roughly nine times better to overcome inertia. Real-world winners like Netflix didn’t win solely on content quality; they won by minimizing friction and making the experience feel like home. The same dynamic plays out with AI tooling: once users internalize a tool’s shortcuts, prompt patterns and quirks, that tacit knowledge becomes a high switching cost that raw model improvements rarely overcome. For AI/ML teams the takeaway is practical and urgent: the primary competitive advantage is interface and experience, not incremental capability gains. Design for an effortless first five minutes, build repeatable interactions that create muscle memory (shortcuts, predictable flows), and let users customize and persist preferences (saved prompts, custom instructions) so leaving becomes costly. Benchmarks matter less to retention than perceived fluency; product roadmaps should prioritize onboarding, interaction models, and data/model affordances that cement user trust and transfer learning — because the model that “feels like home” will likely outcompete the model that’s merely a bit more capable.