What does the prompt actually do? This article claims it "crashes many AI chatbots" and that it "looks easy to DDoS a model at the API level using this prompt",

What does the prompt actually do? This article claims it "crashes many AI chatbots" and that it "looks easy to DDoS a model at the API level using this prompt", suggesting to me that it, for instance, throws some uncaught exception in the tokenizer and brings down the process. The quoted Microsoft and Groq responses ("deficiency in the model's ability to accurately and uniquely respond to the prompts provided", "this is still strange behavior") and the author's description from the previous article ("babble-like madness") imply more that it's a way to induce incoherence in the model responses. That doesn't seem useful at all for DDoSing through the API, where a token costs the same regardless of whether it's coherent and you could already tweak parameters to reduce coherence if you really wanted.