Recently, I had the pleasure of sharing some insights @AI-Engineer's-World-Fair about the work our team at GitHub Next is doing in the space of developer tooling in the age of AI. I truly believe that language models are not the moat; it is the applications that you build around them. Compelling applications often result from designing user experiences that highlight the latent capabilities of these models while restricting or at least moderating their limitations.

Systematic ideation is certainly an excellent way to design and build these compelling and thoughtful experiences. My PhD experience has trained me to do that. However, from my training I also know that these systematic explorations can take a while. Oftentimes, by the time these systematic explorations are completed, the language model capabilities have vastly improved, and more latent capabilities have been discovered.

At GitHub Next, we turbocharge the process slightly by tapping into our intuitions and prototyping capabilities to rapidly iterate on explorations. This is supplemented by having explicit learning goals and ruthless dogfooding to get feedback throughout the prototyping process, ensuring that we are making meaningful progress all along.

I would love to know how you or your team approaches building with AI. Share your insights!