We're a couple from Montreal, not professional developers. We always liked guessing games where you try to say what most people would say, and one day we wondered what the opposite would look like: a real answer that almost nobody would give. That became the game. We built it in about two months. My girlfriend drew the whole look of it: the rabbit in a hot-air balloon, the cliffs, the sky and the abyss.
There are 4 modes. In Sky, you give the answer the most people would give. In Abyss, you give a real answer that almost nobody would give. In Past, it's the oldest answer, and in Future, the most recent one. In Sky and Abyss, the higher you climb or the deeper you dive, the more points you score. In Past and Future, you travel through time instead. There's a leaderboard for the day.
The hardest part, and the most interesting, was understanding answers typed freely, in 10 languages: typos, plurals, nicknames, translations. When an answer isn't on the list, it goes to a "judge" that checks it really exists. The percentages start from an estimate, then move with the answers of real players. If the list missed a good answer, it gets added as soon as a few different players give it. There are also secrets hidden in the game, so go find them.
To be honest about AI: the code was written with AI tools, and without them we couldn't have built this as a two-person team. The questions are generated by AI, then checked and filtered by scripts and a second AI review. My girlfriend's drawings were cleaned up with AI.
We'd love to know two things. Is the goal clear from the very first question? And were any good answers rejected unfairly?