People seem to have forgotten that the point of creating products in small iterations is to get the continuous feedback necessary to produce a great product. It has nothing to do with coding mechanics. The idea is that very few customers know what they want until they get something into their hands. So, get something into their hands, get some feedback, and adjust. Small increments (because nobody, including customers, has the cognitive bandwidth to handle an avalanche of changes all at once), frequent releases (daily is ideal), continuous improvement. Meanwhile, the stans all tell me that the "right way" to use AI is to create a vast amount of code all at once. So much, in fact, that it's not feasible to examine it. Honestly, though, pushing out all that code in one chunk works against us from a product development perspective. I'm not saying we shouldn't use AI, but maybe we're not going about using it in the best way, at least not if we intend to produce products people actually want to buy. That all-at-once mindset has led us back to traditional shove-it-down-the-customers-throats waterfall thinking. Build something huge, some "great idea!" that we're convinced will "disrupt the market!" with its "revolutionary!" approach 🙄. Then try to convince customers that they actually need it. Often, the market remains unconvinced. That's a sales-driven approach. Compare that to a marketing approach, where you figure out what the customers need first, then build that. That's where small increments come into play. It's not a coding strategy; it's a market-research strategy that's more effective (at least faster and cheaper) than doing it with mock-ups. It's a way to figure things out while we're building, rather than before. The traditional critique is that it's somehow "cheaper" to mock up the idea first, but that thinking doesn't factor in the opportunity cost of the time delay and overestimates the cost of actually building it. In this brave new AI world, that cost is lower than ever. The AI advantage is a tight inspect-and-adapt loop that gets us feedback sooner. Leverage that.