Blog
What a learning hackathon is and how our last one went
Hackathon planning

What a learning hackathon is and how our last one went

Inside our 30-day hackathon built around learning a skill, with results from 1,200+ developers.

Text reads "Inside our last Devpost learning hackathon" next to an abstract illustration of a man teaching

The developer population is projected to reach a billion by 2030, driven in large part by agentic software that lets anyone build apps with plain English, not just code. As a result, those developers will want experiences that connect their agents to their workflows and the tools they use every day.

Hackathons have always been one of the most effective ways for developers to experiment, build new skills, and test emerging technology under real conditions. That's a core part of what Devpost does, and something we're constantly thinking about how to push further.

With that in mind, we ran a successful learning hackathon in April: a 30-day challenge built around acquiring a specific skill. In contrast to a regular hackathon, our learning hackathons first teach a new skill, then immediately set participants to work using the skill on a project they care about.

Here's how it went:

  • 1,200+ registrants over 30 days
  • 4.4 / 5 satisfaction rating
  • 80% of participants finished the spec-driven build
  • 98% said it changed how they approach building with AI
"For consulting and hackathon projects, this workflow provides a repeatable framework. It's especially useful when juggling multiple stakeholders or complex integrations." — Participant, 5+ years experience

This post covers:

  • What a learning hackathon is: a format that combines hands-on building with structured learning
  • How we created a Claude Code plugin to guide participants step-by-step through spec-driven development
  • Results from 1,200+ registrants, including what participants said was most valuable and how it impacts their future builds

What is a learning hackathon?

Most hackathons involve challenging participants to build something based on the competition's theme or challenge tracks. Participants have a set amount of time and then their submissions are evaluated by judges.

A learning hackathon is built around a different objective. The goal is for participants to learn something by building. Rather than competing to produce the most impressive project, participants work toward learning a new skill or tool by completing a process.

In our case, the format was:

  1. Install a Claude Code plugin
  2. Use it to plan and build an app of your choosing
  3. Submit

The plugin was an AI learning partner that guided participants through each step from brainstorming to submission. Our intention was for participants to build a process they could repeat. Every artifact produced along the way (scope document, PRD, technical spec, build checklist) was theirs to keep and reuse on future projects.

The entire goal was for builders to learn something new and apply it. Participants of all skill levels could participate.

{{callout-1}}

Inside the Spec-Driven Development Hackathon

We launched the Spec-Driven Development Hackathon and ran it on Devpost for 30 days. Here's a closer look at the framework, including the learning assistant plugin we built to guide participants through their projects.

The curriculum: Spec-driven development

The subject of this hackathon was spec-driven development, which is a way of building with AI that starts with planning.

Many developers using AI coding tools jump straight to code. You describe what you want, the AI starts building, and things generally work until they don't. The context window fills up. You're not sure what half the code does. You start over.

The problem usually isn't the AI. It's that neither you nor the AI had a clear enough picture of what you were building before you started.

Spec-driven development changes the order of operations. Before writing any code, you:

  • Define the scope of your project (including what's explicitly out of scope)
  • Write user stories and acceptance criteria
  • Make architecture decisions and document them
  • Build a checklist for the build phase

With that foundation in place, the AI executes against a plan rather than guessing. The build phase moves faster and produces cleaner code because the thinking is done upfront.

We created a learning assistant plugin for Claude Code

We created a plugin for Claude Code that guided participants through the full spec-driven process. It functioned like a learning assistant that asked questions, pushed back on vague answers, and walked builders through each phase step by step.

The plugin moved participants through four phases before any lines of code were written: scope, PRD (product requirements document), technical spec, and build checklist. Each step was triggered by a slash command, which triggered a new Claude Code skill and guided participants through that step. Each phase produced an artifact that participants kept.

By the time they were ready to submit, every participant had a docs/ folder containing a scope document, PRD, technical spec, build checklist, and working application. Those documents are reusable templates for future projects.

We created detailed resources for builders at every level

For any hackathon, we strongly encourage organizers to provide high-quality resources for participants. We wanted to ensure we put this into practice for the learning hackathon.

We created:

Ultimately, 59% of submitters were beginners or new to coding, but the goal was to make setup clear enough that any participant could get into the plugin and start building quickly, regardless of background.

How it went: Results from the learning hackathon

1,200+ people registered over the 30-day run. There were 130 completed projects submitted. The cohort was 59% beginner-or-below by self-reported experience level.

Results from submitters:

  • 80% completed the full spec-driven workflow
  • 4.4 / 5 mean satisfaction score (57% gave 5 stars)
  • 98% said the hackathon changed how they approach building with AI
  • Beginners rated the experience 4.34 / 5; developers with 5+ years rated it 4.42 / 5

We asked participants for feedback on what they found to be most valuable. They highlighted that writing the spec first helped them with the entire build and shared how they plan to continue using the workflow.

Writing the spec first was the most valuable part of the experience

Most participants said that writing the spec before writing the code was what helped them the most.

"Writing the spec before building. Every component already knew what data it needed and where it came from. The build phase had almost no debugging because the thinking was already done." — Participant, 5+ years experience

A participant new to coding said that defining what was out of scope was even more useful than defining what was in scope. It gave them a clear reason to say no to feature creep mid-build. Another participant said they caught three architecture changes during the planning phase, which cost them minutes, whereas catching them mid-build would have cost hours.

Participants planned to use the workflow in other projects

A significant portion of respondents noted that the workflow felt like something they'd use at work or in other projects.

"For consulting and hackathon projects, this workflow provides a repeatable framework. It's especially useful when juggling multiple stakeholders or complex integrations." — Participant, 5+ years experience

One participant brought the plugin to a separate UCLA hackathon after completing ours and shared how it helped them win a prize:

"I was able to test your plugin at a UCLA hackathon, and I am pleased to announce that we won $500 as an honorable mention."

Multiple submitters wrote that they had started using scope documents and PRDs on their own projects before the hackathon even closed. One was running the process across three active projects.

{{callout-2}}

What's next

This is part of a new series of learning hackathons on Devpost. As the developer population grows toward a billion by 2030, the demand for learning experiences for developers will grow with it. 

Learning hackathons are one way to meet that demand: completable, practical, and built around the tools developers are already using.

But the real magic happens when you look at the bigger picture. Running a hackathon program compounds value across events by providing broader reach, deeper insights into how developers use your tools, and stronger relationships within communities vital to your business.

If you're thinking about how to build a hackathon program that creates lasting value for your developer community, talk to our team.

Get a head start on your next hackathon with our AI hackathon planning guide.
Download now
Thinking of running a hackathon? Talk to our team to get started.
Book a meeting