The video compares three AI models—Clawude 3.7 Sonnet, Gemini 2.5 Pro, and 03 Mini High—by challenging them to create a launch-style game using P5JS, with varying degrees of success. While Clawude fails to meet expectations, Gemini produces a solid game, and 03 Mini High introduces innovative mechanics but deviates from the original prompt.
In the video, the creator tests three top AI models—Clawude 3.7 Sonnet, Gemini 2.5 Pro, and 03 Mini High—by challenging them to create an addictive launch-style game similar to Kitten Cannon using P5JS, without any HTML. The creator specifies preferences for pixelated animals, humorous physics, and random obstacles that can either propel the player forward or halt their progress. The initial attempt with Clawude 3.7 Sonnet results in unexpected behavior, leading the creator to try multiple fixes, but ultimately, the game does not meet expectations.
After the disappointing results with Clawude, the creator moves on to Gemini 2.5 Pro. This model initially presents an error, but after some troubleshooting, the game is functional. However, it encounters a collision error, which the creator resolves with another prompt. The final output from Gemini is deemed significantly better than that of Clawude, showcasing improved gameplay mechanics and overall quality.
Next, the creator tests 03 Mini High, which introduces an interesting launch system. Initially, the creator perceives the launch power as weak, but upon further experimentation, they discover that the launch distance is influenced by how far the player drags the launch mechanism. The game features colored obstacles, with red ones slowing the player down and green ones providing speed boosts. This mechanic adds a layer of strategy to the gameplay.
Despite the unique mechanics of 03 Mini High, the creator notes that it does not fully adhere to the original prompt regarding random obstacles that affect the player’s trajectory. While the game is engaging and offers a fresh take on the launch style, it diverges from the specific requirements set forth at the beginning of the challenge. The creator appreciates the creativity but emphasizes the importance of following the prompt closely.
In conclusion, the video showcases a comparative analysis of the three AI models in generating a game based on specific criteria. Clawude 3.7 Sonnet falls short of expectations, while Gemini 2.5 Pro delivers a solid game experience. 03 Mini High introduces innovative mechanics but strays from the prompt. The creator invites viewers to share their opinions on which AI model performed best, highlighting the ongoing exploration of AI capabilities in game development.