I just released a bunch of updates to Flavorful. The theme, as I wrote last month, was "making Flavorful fun."
I wanted to focus on the idea of fun because cooking improvisationally is fun, so an app that helps you do that needs to be fun too. Planning, shopping, and cooking are a lot of work, no one needs an app that feels like it's adding a bunch more work on top of that.
One idea I tried was a literal turn-based game. I picked an ingredient, then the "computer" added an ingredient, then I added an ingredient, and so on. After each turn the recipe updated. The unpredictability of what the computer would do added some suspense and tension that was pretty compelling. But there was so much randomness to the game that it made me wonder what the goal was in the first place.
Instead, I gave the user and Flavorful distinct roles. As I add ingredients, Flavorful starts suggesting ways they could be combined for me to consider.
For example I start with "roasted green chilies" and Flavorful suggests adding something creamy to get to queso or an enchilada filling. But the decision about what's next is up to me. Feeling puckish, I add "sugar." Flavorful doesn't flinch, and the suggestions get more specific: a spicy simple syrup for drinks, a classic green chile jelly, and candied chile nuts. Turns out my attempt to be mischievous actually unlocked some interesting directions. We are, literally, riffing.
The other push toward fun was adding a daily challenge. Take a look at the cookbooks on your shelf - how many recipes in there have you actually cooked? For most people, it's two recipes. That doesn't diminish the value of the book, or stop people from enjoying cookbooks (or cooking shows, or social media, etc).
So what if a cooking app could be fun enough to use even when you didn't need to cook anything? The Daily Challenge explores this idea. Here's a theme for the day (today it's Allium Family Reunion) and a set of ingredients (onions, shallots, chives, etc). Pick one or pick them all, and see if you can discover something unexpected.
What's next? Well, I've been using Flavorful for the last week and have enjoyed a few unexpected meals (hello, Miso Mushroom Pasta). I'll be reaching out to some folks who have tested previous versions over the next couple weeks to get more feedback. And I've come to the point where I need a more rigorous way to evaluate if the AI model is doing what it should be. Flavorful is currently running GPT-5, which costs about 8ยข per recipe. Switching to a cheaper model, like GPT-5-Mini, would bring those costs down to under a penny, which makes a big difference in thinking about how Flavorful could reach more people. So far I've just been comparing models based on vibes, but now I'll be working on a systematic way to do those comparisons.