To start off 2025, we’re rolling out a company-wide AI integration strategy—and what better way to kick things off than with an AI-driven hackathon? Over a whirlwind 24 hours, our teams leveraged powerful AI tools, including the snarkiest judge you’ve ever heard and the coding superpowers of Augment Code, to bring wild new ideas to life.

Best Practices for an Augmented Hackathon

Before diving into our hackathon highlights, here’s a quick style guide for getting the most out of Augment Code in a time-crunched competition:

Onboard Early

Make sure the Augment Code plugin is installed well before the hackathon begins. Provide a concise tutorial on all of Augment Code’s features so participants have the confidence to dive in and experiment.

Include Technical & Non-Technical Stakeholders

Coding assistance lowers the barrier to entry, allowing product managers, designers, and other staff to participate. Encourage everyone to pitch ideas and get hands-on with the tools.

Use “User Guidelines” for Context

Before you start coding, set up the right context in your codebase. By applying user guidelines, you’re essentially giving your AI helper a map for how and where it should generate code.

Brainstorm Aloud with Your Assistant

Don’t just wait until you’ve decided on a final concept. Chat with Augment Code early during ideation; it’s fantastic at sanity-checking ideas, considering feasibility, and helping you see new angles.

With those best practices in mind, let’s look at how our teams leveraged various AI-driven approaches to build—and occasionally break—things in style.

How We Structured Our AI-Driven Hackathon

A Snarky AI Judge – Because we couldn’t resist coding some personality into the judging process.

5x Development Speed with Augment Code – Turbocharged coding like never before.

An Entire Suite of LLMs – Acting as “rubber ducks” for our engineers, always there to bounce around ideas.

24 Hours from Ideation to Demo – Teams had one day to transform ideas into reality. No pressure, right?

AI Adoption Styles: The Consultant, The Intern, and The Overlord

During the hackathon, we saw a variety of “collaboration styles” between our engineers and AI tools. They fell into three main categories:

1. The Consultant

Best for tackling tough technical challenges with limited resources. Here, AI acted like a seasoned consultant—answering critical questions, surfacing edge cases, and steering the project’s direction without completely taking over.

2. The Intern

When you know exactly what you want but are too lazy (or too strapped for time) to code it yourself, treat AI as your personal intern. Ask for step-by-step help, and refine as you go. Augment Code effortlessly turned prompts into detailed, line-by-line code suggestion

3. The Overlord

When you need a rapid prototype and you’re willing to embrace a bit of chaos, unleash the full power of AI. Let the machine do the heavy lifting and refine with human insight only at the critical moments. Just don’t blame us if the “exponential entropy” creeps in.

Hackathon Highlights

Below are some standout projects that emerged from our 24-hour AI marathon.

Bring Me the Robin

One prompt to rule them all. This project took the notion of “minimum effort” to a new level: an entire codebase generated from a single AI prompt. The result? A tool that instantly empowered our business development team to find the right enterprise contacts to keep expanding AtmosphereTV’s venue fleet.
This was a textbook “Overlord” approach—AI did most of the heavy lifting, while the team validated and tested the result. The resulting code was solid enough to score a perfect 10/10 in the “Ship It Now” category and our business development team loves it.

AI Promo Builder

This project aims to help mom-and-pop venue owners quickly create custom promos to display on Atmosphere TV—even if they’ve never designed a graphic in their lives.
With Augment Code as a handy coding-intern, our product manager built an experimental feature in our existing customer portal in just one day.

“Let’s be real—mom-and-pop shops already have free templates they’re not using. Plus, scraping random websites could be a mess when their niece’s WordPress breaks. But hey, auto-generating content is actually pretty clever.”

Video Sharing Mesh Network

This ambitious project uses a “Consultant” approach to create a cross-platform mesh network in each venue, allowing devices to share video seamlessly. Only one device needs to download the data; the rest stream from a “leader,” enabling high-quality, low-latency playback. Possibly the most technical hackathon submission, the team relied on AI as a pair-programming consultant to tackle complex edge cases—“It seems like it could actually work,” our AI judge said, “unlike half the ‘blockchain for TV’ pitches I’ve seen today.”

Conclusion

In just 24 hours, we developed rapid prototypes—and sometimes production-ready features—by pairing humans with AI tools like Augment Code. This synergy drastically cut development time while giving non-engineers (think product managers, operations, and sales) a chance to shape our software. Yet, as we learned, human oversight is key for larger codebases, and we’ll be establishing AI style guides to standardize best practices. The success of our AI-driven hackathon proves that when you harness AI’s potential—and add a dash of snark—remarkable things can happen fast. Looking ahead, we plan to continue refining these AI-human workflows and exploring new frontiers in AI-assisted innovation at Atmosphere TV.