Brian Balfour

Two Founders Leverage AI Docs, Bake‑offs, Data‑first Design

Kyle Ledbetter and Andy Keil (founders of Dreambase.ai ) have built more product than any other 2 person team I've seen. I first heard them on a podcast a couple of months ago and was fascinated by how they were building. We wrote a full post on their process. There are a few unique things they were doing: ☁️ AIR (AI Requirements Doc) First, they would collapse their entire product vision into one markdown document to feed to AI. It's not traditional documentation; it's specifically designed to make AI effective. Giving it full context on product strategy, user flows, design principles, etc for every conversation. 🧑🍳 Bake Off When it's time to bring that to life, they do a bake off across multiple AI tools to get a variation of ideas. (note, you can do this natively in Reforge Build now) 🧱 Starting with/ The Data Rather than starting with UI, they start with the data schema. This is something Ravi Mehta has also written about and shown how to do in Reforge Build. This creates better outcomes with AI because as Kyle says: "You can't hide behind vague product requirements when you're defining a database schema. What actually is a "user"? How do workspaces relate to projects? Can a user belong to multiple workspaces?" 🙇♂️ Reorganizing Around Learning Last, they split responsibilities based on what type of work cadence is needed to get it done effectively. One is fully dedicated to coding with AI. The other is fully dedicated to anything touching customers (demos, research, feedback, support). They are constantly tightening how one feeds the others. I liked this description on why: "When I'm working with Cursor or Claude, I'm in what I can only describe as a bare-knuckle boxing match with the LLM. I'm close to getting something working. One more prompt. One more test. I know I'm almost there. You can't do that effectively if you're jumping into customer calls every 30 minutes. You need deep, uninterrupted time. " Full post in the comments.