One thing that continues to bother us is simple: we still don’t fully know what kind of experience we want to provide.
We have our target audience — middle school students learning Greek mythology.
We have our characters — Poseidon and Athena.
We have our technology — a three-way conversation system.

But what kind of experience is actually entertaining?

Anyone who has worked with LLMs knows how unpredictable they can be. A one-turn conversation can be amazing; a ten-minute conversation with a middle schooler can go absolutely anywhere. No one can predict the outcome — not even us.

Because of this unpredictability, our core principle hasn’t changed:
different players should receive the same intended experience.

To achieve that, we need strong design and strong control.
This week, we finally started shaping a clear, well-designed journey that lets us map out what we want players to feel, not just what we want them to do.

This ties directly into one of the biggest features we implemented: GeneratedDiscussion.
It allows us to trigger custom AI behaviors based on things like:

• round number
• waiting time
• topic triggers

It’s a powerful tool, basically letting us pre-design “beats” in the experience based on predicted user behavior and what we’ve observed in playtests. It gives us structure inside the chaos.

Art and Animation Progress

We also made major progress on the art side. The animation system is now underway, and we’ll be showing much more next week. This is the first time the gods are starting to “move,” which helps the experience feel more like a performance than a static conversation.

Playtesting With the Real Audience

To bring all of this together, we ran a playtest with our actual target audience — middle school students.
And honestly, the reactions were spectacular.
Seeing their curiosity, confusion, excitement, and occasional chaos gave us confidence we’re moving in the right direction.

Looking Ahead

Next week, we enter our final sprint, aiming to deliver the full experience, from beginning to end, hardware to software. The goal is a complete, playable version that captures everything we’ve been trying to build over the past ten weeks.