Baruch Sadogursky

What if building software didn’t start with code? You’ve got your product docs, user stories, and mocks—all the ideas laid out. But when it’s time to build, what if the next step wasn’t writing code but writing prompts instead?

Imagine this: prompts that describe what the product should do, in plain language everyone on the team can understand. Those prompts turn into tests, and once the tests look right, the AI writes the code to make them pass. Suddenly, everyone—product managers, designers, developers—is on the same page, shaping the product together.

Maybe this is how TDD finally makes sense—not just for developers but for the whole team. Tests are a shared source of truth, driven by prompts, keeping the focus on what we’re building—not just how.

But can prompts really capture intent? What about the AI’s “black-box” code—how do we trust it? And does this shift make building software feel more collaborative—or just different?

In this talk, we’ll explore these ideas, what might work, and what might break along the way. What if building software from prompt to test to code is exactly what we’ve been missing?