The Hacker News discussion revolves around an interactive guide to rate limiting. hxtk questions why rate limiting algorithms don't consider the cost of serving requests or dynamically learn system capacity for fair allocation. He cites Stochastic Fair Blue and Google's Zanzibar system as potential examples but finds a lack of detailed accounts. Several users suggest alternative approaches and technologies. ucarion mentions shuffle-sharding, while jsw proposes an AIMD algorithm with a token bucket for dynamic adjustment. wonnage points to Envoy's latency-based adaptive concurrency and Netflix's implementation. conradludgate promotes the Generic Cell Rate Algorithm (GCRA) for its simplicity and throttling capabilities. The discussion also covers practical considerations. remus highlights the importance of easily understandable rate limits for consumers. smadge shares their experience switching from token bucket to leaky bucket for better concurrency control. mtlynch criticizes the article's presentation and suspects AI generation. Others praise the visualizations and share related resources. The conversation further touches on issues like CGNAT and the trade-offs between different rate limiting algorithms in real-world scaling scenarios.