When GitHub Copilot began auto-completing lines of code — and, not long after, entire code snippets — everyone was wondering: how long until I can describe an app to Copilot in natural language and then just sit back as it is built for me? For the past couple of months, we've witnessed a very good number of experiments inside this space, and GitHub now takes things up a step further as part of the company's annual GitHub Universe conference here in San Francisco: GitHub Spark.
Officially, of course, it's an experiment launched from its GitHub Next labs. However, it allows you to easily build small web apps based on absolutely nothing but your words. Experienced developers can still see and edit the code, and under all that is a GitHub repository, GitHub Actions, and Microsoft's Azure CosmosDB as the default database for applications that need one; it's all optional, though. Ideally, you'll be able to use a chat-like experience to create a prototype and then refine it in subsequent steps.

You then start building the app off of this initial prompt and immediately get a live preview of what it looks like in just a few seconds. You then just go back and forth and refine the app and work with the bot.

"We strongly believe this is the next level of personal software development," GitHub CEO Thomas Dohmke said to me. "These micro apps, they are not meant to be in place of the professional developer. They are meant to be a tool for you to explore ideas, build little bots, little helpers in your day-to-day — or just experiment in software development with natural language."

While GitHub's materials describe Spark as something for building "micro apps," Dohmke said he wasn't quite sure if there were innate limits to how complex an application could get in Spark, and a GitHub spokesperson confirmed that there are no limits on prompts or capacity, at least not currently though that may change after technical preview. Dohmke said that Spark could be used with any web API and leverage AI models of its own, and how thrilling it is to see the extent to which GitHub users will push this tool.

"It will be very exciting to see, honestly, what users can do, how big of an application they can build with just the power of natural language and this simple user interface that really is designed for you to explore ideas, to spark new ideas," he said.

Because GitHub Copilot now lets users choose which large language model they want to use, it's no wonder that Spark offers the same capability: users get the choice between the recent versions of Anthropic's Claude Sonnet and OpenAI's GPT models.

Users can share their Sparks pretty easily, with personalized access controls. Perhaps even more fascinating, though, is the fact that users can then take that shared code and add to it themselves.

Developers wanting to push these kinds of applications even farther can open up the code at will-and change it if that's what's necessary because, admits Dohmke, AI isn't perfect after all. You can, of course, look into the code base, he said. So if you have an understanding of the code base, you can also look at the code directly and modify that, which is often helpful when the AI makes a mistake — which does happen.