Do you know that moment when you realize you’re solving the same problem over and over again? This is where I found myself about a year ago. My name is Noah Flaherty, and I’m the CTO and co-founder of Vellum. After three years of building tools for AI development, I had this crazy idea: What if we built an AI agent that could build other agents?
Seems meta, right? Maybe even a little ridiculous. But the thing is: it really works. And what lessons did we learn along the way? They’re worth sharing, whether you’re an engineer knee-deep in code, a business leader trying to figure out AI adoption, or somewhere in between.
The death of drag and drop (and why it’s okay)
Let me be straight with you: drag and drop is over. I know, I know, we spent two whole years creating these beautiful low-code editors in Vellum. Those workflow diagrams looked great in the demo. Click here, drag there, connect these boxes, and voila, you’ve got an AI system for you.
But what we discovered is this: It’s ugly. error prone. And the moment you need to create something real, something that actually solves complex business problems, those pretty pictures become a nightmare to manage. Any system worth its salt requires a technical understanding that makes the drag-and-drop interface feel like you’re trying to perform surgery with oven mitts.
Think about it. In 2025, we’re all chatting with ChatGPT like it’s our co-worker. We expect that instant gratification, that natural back-and-forth. Why would we go back and forth to click and drag boxes around the screen?
On the other hand, you have a million AI frameworks coming out every other week. Langchain, AutoGPT, CrewAI – pick your poison. But betting your company’s future on a single structure feels like building a house on quicksand. What happens when the next shiny structure falls and everyone jumps ship?
We are stuck in this strange tension. You want to empower more people in your organization to build AI systems, not just engineers. You want Karen from Accounting to automate her workflow and Bob from Sales to create your lead qualification bot. But you also need the robustness and flexibility that only comes from code.
Answer? natural language. It’s an interface we all understand.
For expert advice like this delivered straight to your inbox every other Friday, sign up for a Pro+ membership.
You’ll get 300+ hours of exclusive video content, a complimentary summit ticket, and much more.
So what are you waiting for?
Get Pro+
Tool Definitions: The Secret Sauce No One Talks About
Lesson number one here is, and it might be the most important thing I share today: tool definitions really, really matter.
Prefers traditional software APIs. Computers talk to computers through these clean, structured interfaces. But hey? AI prioritizes conversation. It seeks to interact with systems the same way you and I would interact with them.
Let me give you a concrete example. Let’s say you’re creating an agent that updates Salesforce records. The traditional approach would give your AI three different tools: one to search for contacts, another to get contact details, and a third to update records. This is how APIs work: granular, specific, step by step.
But humans don’t think like this. When I ask you to update a customer’s information, you don’t think “First I will execute a search query, then I will retrieve detailed records, then I will perform an update operation.” You think, “I need to find and update this record.”
So we started abstracting. Instead of three tools, we create one: find_and_update_record. Under the hood, it still performs all three operations. But for AI, it’s just a consistent action. AI thinks like a human, and good old-fashioned code handles the nuances.
Austin’s AI and tech landscape: how it has evolved
Silicon Valley is still at the center of the AI ​​conversation, not because it has a monopoly on ideas, but because many of the forces shaping the future of AI collide here.
