I almost quit an international AI hackathon because I couldn't understand the Vietnamese orientation.
And yet, somehow I ended up becoming the Team Leader and forming a two-person team with a Vietnamese AI engineer.
I wrote about all that chaos in Part 1.
But the part that really surprised me came next.
I shared CSV files I use for my own business. When I met my teammate the next day, he'd already built much of our AI Agent.
When I asked how long it took, he said about three hours!
Wait, you already built that much?!
This is what happened between forming our team, seeing our first prototype, and submitting the Idea Proposal on October 5.
Once Our Two-Person Team Was Set, I Started by Sharing Materials
Our roles were pretty clear: I'd focus on the business side and how the work actually happens, while my teammate would handle most of the technical implementation.
I sell products on Amazon, so I can explain what data I look at every day and what tasks I do.
But I couldn't yet picture exactly how an AI Agent would be built or how it would work.
Since we'd formed the team early, I figured I should share whatever I could right away.
I created a shared Google Drive folder and organized CSV files and other materials from my everyday Amazon work that might help with development.
My teammate replied with one word:
“Awesome.”
Great! Maybe I could be useful after all.
That was about all I was thinking at the time.
At Our Meeting the Next Day, the Agent Was Already Running
We met in person for the first time the next day.
I'd expected us to discuss things like “What should we build?” and “How should we approach it?”
Instead, my teammate had already built much of the Agent and brought it to show me!
Hold on. You already made it?
When did you make it? I asked. He told me he'd done it after looking at the CSV files and other things I'd shared.
It had taken about three hours.
That's ridiculously fast! (Haha.)
Of course, it wasn't completely finished at that point. But I never expected to see something actually working so quickly.
And something else surprised me.
The moment I saw it, I suddenly understood the whole project.
I knew what problems I wanted to solve and what I hoped to do. But I couldn't quite imagine what form that would take.
Watching it run, I thought, Oh! So that's what we're building!
Hearing an explanation and seeing something working are completely different experiences.
Even as a Non-Engineer, I Had Something to Contribute
Honestly, building every technical part of an AI Agent myself would be difficult.
But there are things I know precisely because I work as an Amazon seller every day.
Which CSVs do I use? Where does the time go? What would make my daily work easier?
I shared those details and some actual working materials, and my teammate turned them into something real.
I could leave the technical parts I couldn't do to him, and contribute where I had experience.
With just two of us, that division of roles felt especially clear.
By Our Second Meeting, the Idea Proposal Was Almost Finished
About a week later, we met in person for the second time.
This time we focused on the Idea Proposal for the first round of selection. This proposal itself was the initial judging material. Based on the proposals, roughly 70 teams would be narrowed down to the Top 16.
We'd both been working on the proposal in the meantime, so it was already close to finished when we met.
We were mostly checking the final details together.
If I hadn't seen a working Agent so early on, I absolutely couldn't have written a proposal this specific.
When you can see the real thing, it's much easier to put into words what it does and why it's valuable.
The fact that my teammate built something and showed it to me so early made an enormous difference to me on the business side.
Three A4 Pages Maximum! Harder Than It Sounds
The Idea Proposal had to fit within three A4 pages.
Writing it was hard enough, but the really difficult part was cutting down what we'd already written.
We had so much we wanted to include, but couldn't add more pages.
I suggested, “What if we just make the font smaller and squeeze it all in?” My teammate said no—the formatting was fixed by the rules.
So I thought, What if we translated the English text into Japanese? That might make it shorter!
My teammate agreed that translating English into Japanese can make the text shorter, thanks to kanji.
Then what about Vietnamese? I asked. Apparently that would make it longer.
He explained it like writing Japanese entirely in hiragana, without kanji.
Oh, so kanji is one reason the same thing can take up less space in Japanese! That made sense.
It also made me appreciate kanji. Japanese is amazing. Thinking about it, we did spend so much time practicing all those characters in elementary school!
Discussing the differences between Japanese, English, and Vietnamese while cutting down a proposal—now that's an international hackathon! I loved that part.
Maybe I Could Handle the Visual Side
At our second meeting, I brought up another idea.
I couldn't take on the technical core, so I wanted to leave that to my teammate. But I did want to make the final adjustments to colors, fonts, and the visual side of the UI.
He understood what I meant.
We'd already decided on the name FrostLoop AI.
Once the name was set, I went ahead and made three logo concepts on my own. (Haha.) I showed them to my teammate, and we decided to use the one he liked as the official logo.
Then I made a short MP4 and GIF with that logo, too.
I kept doing this stuff even though I wasn't sure we needed it.
Maybe I just like it!
For my own-brand Amazon products, I also made the logos, product images, product-description videos, and packaging designs myself. I guess I did all that because I enjoyed it, too.
October 5: We Submitted the Idea Proposal in Vietnamese!
The final hurdle was submitting our proposal in Vietnamese.
My teammate said, “I'll share the final version when it's ready, so you can check it.”
Wait—what? I can't read Vietnamese! I told him he didn't need to send it to me.
But I ended up copying it into ChatGPT and checking it anyway. AI really is amazing. I shared the feedback with my teammate.
We revised it, checked again, and made a few more adjustments.
Then, on October 5, 2026, we submitted the Idea Proposal!
Finally! What a relief.
At the start, I was so lost during the Vietnamese orientation that I almost quit.
Now I'd found a teammate, formed a team, seen an Agent running, and made it all the way to submitting our proposal.
We'd come further than I'd expected.
Waiting for the Results. But This Might Not End with the Hackathon
About 70 teams were expected to be narrowed down to the Top 16 in the first selection round.
While we waited, my teammate said he wanted to keep refining the Agent.
Our Idea Proposal's Next Step section even mentioned the possibility of selling the Agent someday.
I started imagining that something originally intended for Amazon sellers in Japan might one day be useful to sellers overseas, too. I wondered whether many of Amazon's standard processes might be similar around the world.
Of course, that was still just an idea.
But I found it exciting to think our hackathon prototype could keep growing.
And I was about to get involved in GitHub collaboration and the look and feel of the project.
Next time: how I worked on the logo and colors—and somehow ended up making a penguin character!
I Ended Up Designing a Penguin for an AI Hackathon! My FrostLoop AI Design Adventure | Part 3
My international AI hackathon adventure isn't over yet.
This article reflects my experience through October 13, 2026. It does not cover the first-round selection results.