文章以冰壶为例说明：AI 正在把「最佳策略」带进休闲领域。阿尔伯塔大学团队用模拟器与机器学习，早在 2016 年就算出「最后一击（hammer）」更优解，按模型建议比奥运等级选手更容易赢；如今高水准队伍与国家队使用分析软体已很普遍。

在可量化的游戏与赌博中，AI 的渗透更直接。扑克研究在 1990 年代与 2000 年代累积，2019 年卡内基美隆的机器人击败多人无限注德州扑克职业选手，之后 AI 既用于训练也取代真人对手。赛马投注方面，比尔·本特在 1980 年代用演算法取代人类直觉，机器学习让系统自动找模式；美国赛场「电脑财团」已占投注金额约 20%–40%，并改变赔率。

作者也提出代价：推荐与最优解可能造成「智识拉平」，把人引向同样的节目、目的地与玩法。扑克因此更内向、数学化，桌上互动变少；NFL 球队雇 AI 专才做策略但比赛中仍被禁止使用。其他例子包括九子棋在 1990 年代被电脑分析出不败策略而变得乏味。即便如此，自动化娱乐需求仍在增长，本特形容「过去一两年」是他「近 20 年」最兴奋的时期。

AI is increasingly optimizing leisure, not just work. Curling shows the shift: University of Alberta researchers built a simulator and machine-learning system to choose the best endgame tactics for the hammer. In 2016 they reported that following the model would win more often than leaving decisions to Olympic-level athletes, and Curling Canada says software analytics are now common in elite teams and national programs.

In games and gambling, the numbers are starker. Poker research progressed through the 1990s and 2000s; in 2019 a Carnegie Mellon bot beat professionals at multiplayer no-limit Texas Hold’em, after which AI became ubiquitous for training and even as opponents. Horse-racing bettor Bill Benter moved from hand-crafted factors in the 1980s to machine learning that finds patterns directly, while computer syndicates account for about 20%–40% of dollars wagered at US tracks, reshaping odds.

The essay argues that optimization can flatten culture and fun: LLMs may steer everyone toward the same shows, trips, and games under a “just for you” veneer. Poker already feels more mathematical and less social, and even when robots are physically weak, strategy in sports is shifting (NFL teams hire AI specialists though in-game use is banned). Past computer “solutions,” like nine men’s morris analysis in the 1990s, can make play less appealing—yet demand for automated fun keeps rising, and Benter calls the past year or two his most exciting in 20 years.