Blueprint

This is the logo for Blueprint

Inspiration

Starting a technical project sounds simple until you're staring at a blank screen asking yourself: What should I build? Where do I start? And what am I actually capable of building?

There are thousands of tutorials, GitHub repositories, project lists, courses, and AI tools available to learners. Ironically, having so many resources can make starting even more overwhelming. Finding an idea that matches your interests and experience is difficult enough; turning that idea into manageable steps and actually finishing it is another challenge entirely.

We also noticed a shift in the way generative AI is used for programming. Many AI tools are optimized to generate code as quickly as possible. That's incredibly useful—but for someone trying to learn, having AI build everything for you can take away the most important part of a project: understanding how to build it yourself.

That led us to a question:

What if AI didn't build your project for you, but helped you learn how to build it?

Blueprint was created to bridge that gap. We wanted one place where learners could discover projects, turn their own ideas into actionable plans, receive guidance when they're stuck, track their progress, and eventually showcase what they've created.

Instead of handing you the finished product, Blueprint gives you the blueprint.

What it does

Blueprint is an AI-powered project discovery and learning platform that helps people go from "I want to build something" to "I built it."

Users begin by telling Blueprint about their technical interests, skills, and experience level. Whether they're interested in software engineering, data science, cybersecurity, web development, or another area of technology, Blueprint uses that information to make project discovery less overwhelming.

From there, users have two ways to start.

Discover a project

Blueprint recommends existing open-source projects based on the user's interests and experience.

We collected GitHub repositories across different technical areas and evaluated repositories using signals such as GitHub stars to prioritize established projects. Instead of dropping users directly into an unfamiliar repository, Blueprint transforms that information into readable project cards that make it easier to understand what a project is, what technologies it uses, and whether it matches the learner's interests.

Users can still access the original GitHub repository when they're ready to dive into the technical details.

Plan your own idea

Already have an idea?

Users can describe what they want to build and use our Gemini-powered planning experience to transform that idea into a structured project.

Blueprint can help organize the idea into a recommended technology stack, project tasks, milestones, and timeline displayed through a project dashboard.

The goal isn't for the LLM to generate the entire project. We designed the AI experience around guidance rather than replacement: helping users understand their next step while keeping them responsible for actually building it.

Get help while building

Blueprint also includes an AI mentor designed specifically around the learning process.

Rather than behaving like another "generate my entire app" chatbot, the mentor is prompted to help users work through their projects. It can provide guidance, explain concepts, and help users determine what to work on next.

We also incorporated voice interaction using ElevenLabs, giving users another way to communicate with their mentor while they're building.

Track and showcase your work

Starting a project is only half the challenge. Finishing one requires keeping track of what you've done and what still needs to happen.

Blueprint gives each project its own dashboard so users can follow their tasks and progress throughout the build.

Once they've created something they're proud of, their projects can live on their Blueprint portfolio. Instead of separating project discovery, project management, learning, and portfolio building across several platforms, Blueprint brings them together into one experience.

How we built it

We built Blueprint as a web application using Next.js, TypeScript, and Tailwind CSS.

The frontend was first designed and prototyped in Figma, where we developed Blueprint's visual identity and user experience. Because this hackathon had a heist theme, we wanted the application to feel like users were preparing for a mission rather than filling out another learning dashboard.

That idea influenced everything from our project cards to our interface and ultimately reinforced the Blueprint name: every project is a mission, and every mission needs a blueprint.

For project discovery, we used the GitHub API and web scraping to collect repositories across several areas of computer science and technology. We used repository signals such as stars to help identify reputable projects, then transformed that information into structured data that could populate reusable project-card components.

For AI-powered project planning and mentorship, we integrated the Gemini API. Rather than using a completely open-ended prompt, we designed the model around a specific mentor role and structured its behavior so its responses could support the rest of the Blueprint interface.

This allows information generated from a user's project idea to become part of their project dashboard instead of remaining trapped inside a chatbot conversation.

We also integrated ElevenLabs to bring voice capabilities to the mentor experience, giving learners the option to interact with Blueprint more conversationally.

Our application and services were deployed using Vultr, while GitHub served as both part of our development workflow and an important source for Blueprint's project-discovery system.

Challenges we ran into

Building Blueprint involved almost as much iteration as the projects we're trying to help people build.

One of our biggest technical challenges was integrating the Gemini API with the rest of the application. Getting an LLM to return an answer is one problem; getting it to reliably behave like a learning-focused mentor and produce information that can actually power an application is another.

We had to think carefully about how the model should respond, what information it needed, and how its output should interact with the project dashboard.

We also ran into the classic hackathon enemy: merge conflicts.

With multiple people developing different parts of the application simultaneously, we learned very quickly that Git doesn't replace communication. Changes sometimes reached main before everyone understood what was being modified, creating conflicts and integration problems that forced us to become much more deliberate about communicating before merging.

Design was another major challenge. Our original interface worked, but it didn't feel like Blueprint. About five hours into development, we made the difficult decision to scrap much of the original direction and redesign the application around a stronger visual concept that better fit both our project and the hackathon's heist theme.

We also had to balance different ideas about Blueprint's scope. There were countless features we wanted to add, but with limited time, we had to decide what actually supported our core experience and what needed to wait.

And, of course, after pushing changes to main, we experienced the moment every hackathon team loves: our server went down.

Debugging, redeploying, merging, redesigning, and integrating APIs under a deadline became just as much a part of Blueprint as writing the actual features.

Accomplishments that we're proud of

We're proud that Blueprint became more than a project recommendation page.

We built an experience that connects several parts of the project-building process that are normally separated across different platforms.

Our project-discovery system pulls real GitHub repositories across multiple technical fields and transforms them into approachable recommendations for learners. Instead of requiring someone to already know exactly what to search for on GitHub, Blueprint helps them discover projects based on what they're interested in learning.

We're also proud of our AI mentor and project-planning experience. Integrating Gemini wasn't just about adding a chatbot to the website—we designed the LLM around the idea that AI should support the learner without replacing the learner.

We connected that experience with project dashboards so AI-generated guidance can become actionable steps instead of disappearing into a conversation.

We incorporated ElevenLabs to give that mentor a voice, allowing users to interact with their project guidance in another format.

And we're especially proud of Blueprint's design. After completely reconsidering our original direction during the hackathon, we created an interactive interface that feels connected to both the project's identity and the heist theme without sacrificing usability.

Most importantly, we built something we would genuinely want when starting a new technical project ourselves.

What we learned

One of our biggest lessons was that having access to information isn't the same as knowing what to do with it.

GitHub already contains millions of projects. AI can already generate project ideas. Tutorials already exist for almost every programming language and framework.

The problem we became interested in wasn't simply giving learners more resources.

It was helping them navigate those resources.

We also learned that designing an educational AI experience requires a different mindset from designing an AI coding tool. The fastest answer isn't necessarily the best answer for someone who is learning. Sometimes the better system gives you a hint, explains why something works, or points you toward the next step instead of immediately producing the solution.

Technically, we learned a lot about working with LLM APIs, structuring model outputs for an application, integrating voice AI, working with GitHub data, deploying our application, and coordinating development across a team under a very short deadline.

We also learned firsthand how important communication, version control, task delegation, and scope management become when several people are building one product simultaneously.

What's next for Blueprint

Blueprint started as a hackathon project, but there are several directions we'd love to continue exploring.

First, we want to expand the range and diversity of projects available through the discovery system. More technical disciplines, difficulty levels, technologies, and project types would allow Blueprint to support learners with very different goals.

We'd also like to integrate courses, certifications, documentation, and other learning resources directly into project roadmaps. If a project requires a technology a user hasn't learned yet, Blueprint could do more than identify the gap—it could help them find a path to learn it.

Another direction is making project building more social and motivating. Streaks, milestones, achievements, friendly competition, and collaborative projects could give users additional reasons to keep building instead of abandoning projects halfway through.

We also want to continue developing the AI mentor so it can better understand a learner's existing skills, previous projects, current progress, and learning goals while still maintaining our core philosophy:

Don't build it for me. Help me learn how to build it.

Ultimately, we want Blueprint to become the place where someone can arrive with nothing more than "I want to build something" and leave with the skills, structure, experience, and finished project to prove that they did.