How I lost a month to AI handoffs, what I changed, and what’s coming to Classic and beyond.
My last Cranky’s Game post was September 7, when I introduced the real three-legged turtle behind the game. If you’ve wondered why I’ve been quiet since then, the answer is a little ridiculous.
I’ve been working on the game. I’ve also spent a remarkable amount of time managing the AIs that were supposed to help me build the game. 😄
When AI handoffs needed more helpers
One of the fascinating things about building Cranky’s Game with AI is how much one person can attempt without assembling a traditional development team. The other side of that coin? It is entirely possible to create your own miniature software bureaucracy, staffed by bots.
I had divided the project among five specialized GPTs, each responsible for part of the work. That sounded sensible: keep the responsibilities clear, document everything, and hand work from one area to the next.
But the AI handoffs gradually became the work. A GPT needed a worker to package an update. The package went back to the GPT, which then needed another worker to unpack or process what the previous worker had just delivered. Sometimes it felt like I was commissioning a project just to move the results of the last project to the next desk.
Too many workers. Not enough turtles.
I was still tracking progress, and there was real progress. The problem wasn’t that everything had stopped working. It was all the little side quests: another tiny fix, another verification step, another handoff, another procedural question. Some checks absolutely matter, especially in a game built for kids. But a lot of the smaller issues weren’t actually preventing us from moving the game forward, and they were consuming time I didn’t have much of.
Somewhere in there, I lost the better part of a month to overhead.
Resetting AI handoffs: one team, one direction
So I changed how I’m running the project and handling AI handoffs.
Instead of five GPTs passing work back and forth, I’ve consolidated the main development effort under one ChatGPT-led workflow. ChatGPT coordinates the work, and ChatGPT Work handles the hands-on development and initial testing. Claude has a separate, important role: independent technical and security review at meaningful checkpoints.
That gives me something much simpler: one place to drive the work, another set of eyes to challenge it, and me making the product decisions. We’re not abandoning testing or safety. We’re trying to spend our effort on the tests and decisions that actually help us ship a better game.
And yes, I still get to be the guy asking why the turtle isn’t where it’s supposed to be.
Where Cranky’s Game stands right now
We’re back on track, with a clear sequence for finishing the Classic experience. The immediate focus is getting Visits and Chat right—not just technically working, but feeling natural to a child playing the game.
When a friend visits your sanctuary, you should see their turtle, with their own appearance, not a copy of yours. You should be able to talk right there in the sanctuary, move between places, and have the visit end cleanly. A separate Chat page shouldn’t feel like the center of a game that’s supposed to be about exploring together.
That specific visitor-appearance issue is being corrected and tested in our development environment now. I’m not calling it finished until the two-player tests and independent review support that.
What’s coming to Classic
I’ve also decided we’re not moving on from Classic the moment two turtles can stand next to each other and exchange messages. Friends need things to do together. Here’s the order we’re working toward:
- Visits and sanctuary Chat: recognize your friends, see their actual turtles, and talk while exploring.
- Player interactions: wave, cheer, laugh, follow a friend, and interact with the world together.
- Parties: invite friends into a sanctuary, gather, and start shared activities without leaving the world behind.
- Cooperative mazes: adventures designed for one to four players, with shared progress, puzzles, and something worth celebrating at the finish.
- Interactive exploration: scavenger hunts, discoveries, hidden surprises, and little reasons to wander the Woods, Pond, Garden, and Meadow together.
- Rewards and decorating: earn keepsakes or decorations through play and bring them home to your sanctuary.
That’s the roadmap, not a claim that all six are available today. The goal is to get Classic to a genuinely fun multiplayer stopping point, with working features, real testing, and my sign-off—not to keep building menus just because we can.
And then… the new look
I haven’t forgotten about the next generation of Cranky’s Game graphics. We’ve already explored a more immersive Living Sanctuary look, including day and night artwork and a more expressive world. Once Classic’s multiplayer foundation is solid, that updated visual experience becomes the next major focus.
The important part is that I don’t want to build two separate games. Friendships, visits, Chat, parties, adventures, and the family/privacy protections should be reusable underneath, even when the world looks very different on top.
Think of Classic as getting the game’s heart and hands working together. The updated graphics are how we’ll let that world really shine.
Still the same reason I started
The funny part is that none of this started as an experiment in AI project management. It started because my daughter wanted a place where she could care for Cranky again.
I still want Cranky’s Game to be a gentle, fun world: no pressure to log in every day, no need to spend money just to enjoy it, and no forgetting that real families are behind the players. Kids should be able to make friends, explore, and have a good time. Parents should have a meaningful place in that experience.
I don’t have a release date to announce here. What I do have is a clearer plan and a much better way to spend the hours I can give this project.
Less AI paperwork. More turtles going on adventures.
That feels like progress. 🐢