Show HN: Shuttle AI – Instant apps, from a single prompt (shuttle.rs)

Hey HN!

I'm Christos, one of the co-founders at Shuttle[0], and I'm really excited to share some tech we've been working on over the past few months - Shuttle AI, a tool to generate and deploy fully working Rust backends & services from a single prompt.

Looking at the latest advancements in the LLM field, we started wondering about how we can enable the machines to do more for developers - and as a company that’s focused on developer productivity, we found that our infrastructure-from-code stack could be the glue the machines need to not just generate code, but also take a step forward and deploy it, enabling them build a fully-running service from scratch.

Along with our large repository of sample code, we felt like we were in a perfect position to try and pull off an experiment - and today, we’re happy to show you the results of our work!

With typing just a simple prompt such as `shuttle-ai build “Build me a blog service”` or `shuttle-ai build "Create a simple twitter clone”`, you can have a whole service generated and running in minutes. We’ve developed multiple GPT Agents that operate in sync, starting from breaking down your prompt into a specification, generating the necessary code, error checking it, provisioning infrastructure, and ending up with a running cloud deployment.

While it still comes with limitations, mostly imposed due to underlying models, we are actively working around some of them and working on adding new features to test the limits of the LLM code generation. We’ll be releasing a larger write-up soon around methods involved, limitations we’ve encountered, and how we managed to navigate around them along with some goodies such as `shuttle-ai add-feature "Comments feature for my blog"`.

To see Shuttle AI in action, head on over to our website where you can also sign-up for the waitlist: https://shuttle.rs/ai

[0] https://www.shuttle.rs/