Core Insights - The article discusses how a father, KiddFlash42, utilized AI to create an educational game with his 5-year-old son, transforming the learning experience into a fun activity [2][12]. Group 1: Game Development Process - The game development process was driven by the child's imagination, with minimal documentation and direct communication with AI for coding [4][6]. - The father initially faced challenges with code organization and technical issues, which were addressed by switching to GitHub Copilot, significantly increasing productivity [7][5]. Group 2: AI Performance Evaluation - KiddFlash42 ranked various AI models based on their performance, with Claude Opus 4.5 being the most effective, followed by Gemini and GPT-5.1 [9][10][11]. - The AI models had different strengths and weaknesses, impacting the overall development experience and project stability [10][11]. Group 3: Final Outcome - The final product was a playable educational game that integrated reading, spelling, and math into engaging gameplay, leading to a shift in the father's approach to his child's learning [12][13]. - The father expressed a sense of accomplishment, feeling more involved in the development process despite lacking programming skills [14].
一行代码都不会，花270元、烧光1500次请求，他和5岁儿子一周做出游戏：现在作业直接“玩上瘾”了……
3 6 Ke·2026-01-15 13:19