Day of AI Australia Program Director Natasha Banks speaks with developers Oliver Xu and Joel Paul about their work as game developers with Day of AI Australia, what they learned from students and teachers, and why every design decision starts with the student. The development of several new games was possible with support from Google.org’s Generative AI Accelerator funding, provided to Day of AI Australia and UNSW Sydney.
When we set out to build a suite of interactive games for Day of AI Australia’s 2026 lessons, we knew we needed developers who could do more than write good code. We needed people who genuinely cared about young learners, who could hold safety and creativity in equal measure, and who were willing to let students and teachers lead the process.
We found that in Oliver Xu and Joel Paul.
Between them, they have built several new games now available on Day of AI Australia’s Game Studios platform, with two games launching this week: Sunny’s Mindful AI Day for the AI and Learning lesson for Years 7-10, and Win the Farm for the AI and Elections lesson for Years 7-10.
The games are not supplementary resources. They are immersive, carefully crafted experiences designed to have students critically engage with some really important questions: How does AI shape what we believe? What happens when we outsource our thinking?
I sat down with both of them to find out how they got here, what they built, and what the kids taught them along the way.
Natasha: Let’s start at the beginning. What drew each of you to technology, and how did you end up working on something like this?
Joel: I have been interested in computers and technology since I was a kid. I loved discovering Flash games on the web, and that inspired me to teach myself how to code while I was still at school. I made mods for Minecraft, a few simple games of my own. Maths and science were my favourite subjects, so studying Computer Science and Physics at UNSW felt like a natural fit.
Oliver: I have always had a fascination with learning about how things work, from building Lego to taking an interest in cars, trains and planes. So eventually, I ended up studying aerospace engineering and computer science at UNSW. Throughout uni I worked on many interesting projects. The highlight was being able to build software powering a total artificial heart. I have also had the privilege of being a tutor at UNSW for five years, helping inspire the next generation of engineers.
Natasha: Tell us about the games. What has each one been designed to teach, and how does the game mechanic serve that goal?
Oliver: In Sunny’s Mindful AI Day, students learn about the effects of cognitive debt through overusing AI. The game follows Sunny through her daily life, presenting scenarios that all school students would have experienced. Students have to decide whether Sunny uses AI for these tasks or find out if there is a better method. Through making these choices, students reflect on their own AI use and discover how they can be more mindful users of AI.
Joel: Win the Farm was a national competition for Year 7 to 10 students that took place during Media Literacy Week in 2025. The game’s story is loosely based on George Orwell’s Animal Farm, where the tyrannical pig Napoleon rules over the farm. Our game introduces an opposing leader, Hetty the Hen, who thinks she can do better, and so they decide to have an election to determine the future of the farm. The 2025 game was great for a competition, we needed a version of the game that was available all year, so teachers could deliver the AI and Elections lesson when it works for them. This year I’ve been working on this new, single-player version of the game that condenses the two-week competition down to about ten minutes.
The animals are simulated through LLMs on a fictional social media site called Barn Wall, where they discuss opinions and debate each other – just like in the real world. Students are given a candidate to work for and tasked with gaining votes by creating social media bots that influence the animals’ opinions. They are able to configure the bot’s personality, subject focus, and whether it should be truthful or spread disinformation. This hands-on experience gives them a real feel for how social media can shape people’s thoughts and opinions, and confronts them with the question of whether there is actually a person behind the keyboard. I hope students come away being more critical of the information they see online.
Natasha: Safety is something we are deeply committed to at Day of AI Australia. How did you approach that in the build?
Oliver: For all our games, we followed the eSafety Commissioner’s Safety by Design guidelines to ensure our games were safely designed from the ground up. This meant things like never allowing students to have free text input directly into an LLM. We also had to implement strict content filtering processes to ensure that all content generated in the game by LLMs is age-appropriate and safe. We used Google’s ShieldGemma model, which evaluates content with strict policies and filters out any harmful content.
Joel: For the competition version of Win the Farm, the login system was an important piece of the safety architecture. Teachers log in using their email and a one-time password sent to their email to access their dashboard. The actual logic and code was not too complicated, but it was my first time using AWS (Amazon Web Services), so it was a bit confusing getting everything set up and finally working.
The scale of Win the Farm made safety even more critical. Our platform had to support hundreds of teams, which created nearly 1,000 bots in total running simultaneously. These bots generated over 100,000 posts, and all of this had to run without a single performance issue. We were fortunate enough to be able to make use of Google Cloud’s auto-scaling infrastructure, which was initially quite tricky to set up but solved much of the scalability goals.
For the full story of how we built the infrastructure behind Win the Farm competition version for 2025, the Google Cloud case study is worth a read: cloud.google.com/customers/unswdayofaiaustralia
Natasha: You both spent real time in user testing with teachers and students. What did that process look like?
Oliver: User testing played quite an important role in the development of all our games. We recently ran a couple of sessions with some primary and secondary students to hear their thoughts on Sunny’s Mindful AI Day. These sessions typically involved presenting a small section of the lesson material that the game ties into, followed by the students trying out the game. We gather feedback through a survey and prompt some discussion during the session too.
Joel: For the 2025 competition version of Win the Farm, we held in-person testing sessions at UNSW with students from several schools. They were first introduced to the concepts of misinformation and disinformation, media literacy, elections, and the use of AI with respect to this. We then introduced Win the Farm to them and asked them to try it out. The user testing was really helpful. It was clear that a lot of students were getting confused at some parts and the game needed to communicate what was happening at any given moment a lot better. It was great to see them getting engaged by what we made, asking questions, providing feedback, and bringing new ideas. We similarly tested the singleplayer version of the game with students, noting how they interacted with the game’s elements and reviewing their feedback. Since this was built off of the competition version, the major mechanics of the game were received quite well. The main complaints were to do with players feeling lost on what to do and the amount of information they had to take in. The student and teacher feedback helped us streamline and condense the game to create something approachable and engaging.
Natasha: Was there a moment in testing where a student did something that surprised you or shifted how you were thinking about the game?
Joel: In one of our test sessions, the students picked Napoleon the pig as their candidate and were in the process of creating a bot. When it came to selecting their profile picture, they decided to pick the chick, as they reasoned it would appeal to the hens on the opposing side. At that point in time the animal you picked had no impact on the simulation – it was purely cosmetic – but I was delighted by their thought process. As a result of this interaction, I decided to implement it into the game for real.
Oliver: During one of our user testing sessions with Sunny’s Day, we wanted to gauge how the main character’s movement through her world could work. It was interesting to see how teachers preferred not to have movement (that students would go through a set experience in the game), as it could potentially shift the gameplay from making choices to more exploration, which could distract students. We took this feedback into account and elected to have some movement, but with guidelines to ensure students focus on making choices as they follow Sunny through her daily life.
Natasha: Did teachers and students engage with the games differently?
Joel: Students tended to engage with the content at face value, for example “This game is about the use of AI in social media”, whereas teachers tended to see more of the underlying themes and potential, for example “This is not just useful for Technology lessons, it is also applicable to English and Humanities classes!”
Oliver: Teachers naturally focused more on the pedagogical value of the game itself, whereas students were more interested in the gameplay and the fun it brings. Both kinds of feedback are essential. If the learning is not there, a teacher will not bring it into the classroom. If it is not engaging, students will not lean in. You need both.
Natasha: What has building for education taught you that you did not know when you started?
Oliver: Building for a wide range of ages and reading levels has definitely changed my approach to designing software. I had to adapt to a younger audience, which meant focusing much more on accessibility and readability, and ensuring that all game mechanics are as intuitive as possible.
Joel:
We need to make sure every student is able to engage with our game. That means every button and interaction must use accessible language and be thoroughly explained, but also those who wish to delve deeper into the topic should also be able to do so.
Natasha: Inclusion is core to our mission at Day of AI Australia. How have you approached making these games accessible to all Australian students?
Oliver: We have a strong focus on making our games as inclusive as possible, with the goal of encouraging girls and First Nations students to explore AI through these games. We aim to have female characters represented as much as possible in our games, for example Hetty the Hen and Sunny as lead characters so girls can see themselves in the games, and indigenous artwork is featured in Inky’s Great Adventure. We also aim to ensure that our games are able to reach rural and regional areas, and government school kids in particular. This means building worlds and scenarios that are relatable for all kids in the country, as well as ensuring that our games run on all devices without any performance issues.
Joel: With Napoleon as a male candidate, having Hetty be an opposing female candidate was a thematically appropriate choice. We also try to introduce diverse name options for the bots, as well as using language and concepts appropriate to Year 7 to 10 ages.
Natasha: What are you most proud of in what you have shipped?
Joel: I am familiar enough with the project that I tend to look past its strengths and focus on its flaws, so seeing students actually get excited by the game and engage with it enthusiastically is a great feeling. It is a reminder that what we are building will be experienced fresh by students who have never seen it before, and that perspective matters.
Oliver: For me it is the safety commitment to safety. We hosted more than 100,000 AI-generated posts through the 2025 Win the Farm competition and did not have a single safety incident. That does not happen by accident. It is the result of every design decision we made from day one.
The games Oliver and Joel have built, are now available in Day of AI Australia lessons for Years 7-10.
The games are also available at Day of AI Australia’s Game Studios platform at studios.dayofaiaustralia.com.
Whether you are a teacher looking to bring AI literacy into your classroom, or a parent curious about how your child can engage with these ideas in a safe and structured way, we would love for you to explore.
And if you want to go deeper on the technology behind Win the Farm and how Google Cloud helped us build a safe AI simulation for hundreds of thousands of students, the full case study is at cloud.google.com/customers/unswdayofaiaustralia.
Oliver Xu and Joel Paul are developers at Day of AI Australia, working in partnership with UNSW Sydney. Dr Jake Renzella is the educational game lead.
Sunny’s Mindful AI Day and Win the Farm were developed with funding Day of AI Australia and UNSW Sydney received from Google.org as part of the 2025 GenAI Accelerator.
Bring AI literacy to your classroom with Day of AI Australia
Our AI literacy program gives students the knowledge and critical thinking skills to engage with AI more thoughtfully.
It’s free, curriculum-aligned, and designed to be taught without a computer science background.
Sunny’s Mindful AI Day can be found in Years 7-10 AI and Learning lesson.
The single player version of Win the Farm is now included in the Years 7-10 Elections and AI lesson.
