AI will happily design the wrong thing for you

https://news.ycombinator.com/rss Hits: 2
Summary

I need to clear something up about my book, “Products People Actually Want.” When I write about how “anyone can build anything” now, some people assume I’m anti-AI. That I think these tools are ruining design or product development. That’s not it at all. AI tools are incredible leverage. They make me think faster, broader, and help me produce better work. But like any creative partner or colleague, this doesn’t happen out of the gate. I’ve spent hours—days—training my AI on how we write at Summer Health so the tone is right. I’ve taught it our business model. I tweak its suggestions to better match the experience we want to build. The problem isn’t that AI exists. The problem is how most people use it. The real issue with “anyone can build anything” My book focuses on a specific problem: people building things without knowing if anyone actually needs them. When the barrier to building drops to zero, we get flooded with products that work fine but solve problems that don’t exist. But there’s a second issue too. While AI can help you get started quickly, it won’t build something ready for mass market. Users’ expectations for polish and detail are much higher than what AI produces by default. You can spot AI-generated work from a mile away because it lacks the intentional decisions that make products feel right. The combination is brutal: people building the wrong things, and building them poorly. Just last week, we saw a perfect example. Food influencer Molly Baz discovered that Shopify was selling a website template featuring what she called “a sicko AI version of me.” The image—a woman in a red sweatshirt eating an onion ring in a butter-yellow kitchen—looked almost identical to her cookbook cover. This isn’t really an AI problem, it’s a laziness problem that AI makes more tempting. What used to be someone using an image without permission now gets dressed up as “AI-generated content.” The ironic part? Creating something unique with AI tools is actually easier than ...

First seen: 2025-09-30 16:38

Last seen: 2025-09-30 17:38