GitHub just launched Spark and I cannot wait to tell you more about it. My mind is blown. I’m not even exaggerating here. I literally built a working full stack application in just a few minutes with this thing. And you are going to read an entire demo in just a few moments. But first, let me quickly explain what Spark actually is.
What is GitHub Spark?
So, we’ve been seeing all kinds of AI coding tools lately on the market, right? Cursor, Replet, etc. Well, GitHub said, “No, let’s skip coding altogether.” Spark is basically an AI powered app builder that runs entirely in your browser. But here’s the crazy part. It is not just generating front-end mockups. We are talking about full stack applications with:
- Database authentication
- AI features
- And the whole nine yards.
It’s powered by Claude Sonnet 4 and it has its natural language editor where you literally just describe what you want to be built. Then it has its own runtime environment that hosts everything, persistent storage, and you can even add other large language model features without touching any APIs. Plus, it’s got this progressive web app dashboard, so you can build, test, and deploy from literally anywhere. No local setup, no configuration hell, just pure browserbased development.
Look, I had to test this thing myself because hey, it really sounds amazing. So, let me show you exactly what happened when I gave GitHub Spark a try.
Putting Spark to the Test
The Spark interface allows you to type in natural language exactly what you want to build, and it will build you a full stack application.
Now, if you’re like me, you would want to create a really good prompt. And so, I went into Claude. I gave it my idea, asked it to create a PRD, a product requirements document for a personal nutrition and fitness app. I gave it a few little descriptions of what I want. I also asked it to connect to Gemini flashlight API for taking images of the food so that you can extract macros. It did an absolutely amazing job trying to take all of my requirements from natural language and converting it into a PRD document. I was looking for Gemini. It did actually put my description for Gemini and for data management into the prompt.
After I was satisfied with this prompt for the PRD, I went into Spark and then pasted that prompt as is. Now that I was happy with it, I was able to click on submit prompt. And as soon as I did that, Spark went into its think mode. On the right, I absolutely love how it changes these little status messages, calling itself “I’m building tiny things,” “consulting,” “researching, blockchain buzzwords.” I’m just enjoying reading the different nuggets that they have on the on the right hand side while on the left it is actually trying to create my code.
In just a few minutes (I did a little bit of a speed up here), it had my login information for profile and logging in, and it was taking all the details that I gave it in the description. However, on the quick actions, it was not doing a good job with scanning the food and manual entry. I don’t think it actually connected itself to the Gemini API. So, we’ll have to go fix that and figure out how.
But on the left in the themes section, you’ll see that I could actually change the appearance pretty easily, which I really, really love. And I don’t have to worry about writing code. So this is really for people who want to do this in natural language. Then I’ve got the prompts and the assets and the data. On the right, I was able to see that I can go into the mobile mode to see how it’s going to look on mobile, and convert it back.
Fixing the LLM Connection
Now, let’s go actually fix the problem with our log food feature. I created another prompt and I was really asking it to figure out what’s going on here and make sure that you connect to Gemini Flashlight. And really what I was trying to test here is Spark’s ability to connect to other large language models. I kind of sped things up here a little bit as well to try and see. It was creating a settings dashboard. And there we have it. According to it, it has connected to the Gemini API.
Now, I don’t know if it has done that. It gives you prompts on the left as a follow on, so it kind of knows what you’re trying to do. I asked it to actually go set up the Gemini API key for me, and I saw in the settings that it did that, but it was also requiring the API key from me. So, something somewhere did not go right. I went into the settings myself and then I tried to see where the Gemini API key was, and then I manually pasted it myself.
I asked it to run the app again and see what happened, and it went ahead and did that. And now at this point, I was actually able to upload an image or take a photo. I love Chola Batura. I uploaded that and wanted it to analyze Chola Batura into the different macros that this food has. Then I went in and tried to do the same in the text description. It analyzed that text and created the macro profile for me, and it added all of that and tracked my today’s progress.
From App to Repo to Live Site
On the top, I could go in and create a repository of this code. So just two button clicks and I had a repository built, which is amazing. I was seeing the exact same thing in my mobile app as well. And then when I wanted to go to the repository, I was able to see the GitHub repo. Now I can use or repurpose this repo in any way, shape, or form. I can also open the code in CodeSpaces. So, CodeSpaces, if you don’t know, is basically a hosted version of VS Code and it allows you—it’s so easy to use that it is browser-based and you can very easily go ahead and create those or edit anything that you want to in this code. So that’s a really good thing to have.
And then in the settings section here, there is just a little bit of a description of what’s going on. And with the publish button here, I was actually able to publish this application entirely just in one click. There I had my app live. It was only visible to me right now. So I could go in and see this application live and how it’s performing. This is great. I love it. Oh, I could see my API key as well. I could test if it’s working. Really good experience here. Now I could actually make it visible to all the GitHub users. Again, I don’t really care right now, so I was just giving it a try and it worked really well.
Adding Features Post-Deployment
Now on the left in the suggestions section, let’s say now we’ve published the app. I wanted to see what the experience is like after adding an additional feature. So I asked it to add some social feed feature into this. And after thinking for maybe like a minute or two (I’ve speeded this process up here for us to get to it faster), once it was able to create all of the requirements for the social page that I asked it to do. All right. Once it had done that, I saw that at the very top of the page, it had added a social section, and that is what I wanted.
Okay. So now let’s see what happens. When I went to create a repository, it was asking me to create a new repository. That is not right, because we actually just edited it. But anyway, it actually took the right repo when I clicked on it and updated that repo, which is perfect. And then I had this update button which actually allowed me to update the application that is already live. This is amazing. I’m again not really writing any code. It’s doing all of this for me and it is updating this application with the new social feature that we added after deploying it once. And when I clicked on the website, I was actually able to see my social feature. Great. This is absolutely mind-blowing.
The Honest Take
Okay. So, after seeing Spark in action, what is my honest take? Well, it is a paradigm shift. We are living in an era where we are democratizing software development in real time. Is it perfect? Absolutely not. Complex enterprise applications still need that traditional development approach. The AI can make mistakes and you still need to understand software development principles to truly build applications for production. But for prototyping, internal tools, personal projects, and even early-stage SaaS products like startups, Spark is an absolute game-changer. The speed from idea to working prototype is mind-blowing.
Right now, Spark is available in public preview for GitHub Copilot Pro Plus subscribers at $39 per month.