from Hacker News

It's got my contact from a thematic group I'm in. I can see that the messages are LLM-generated. It wants me to either visit a website or share personal data.

The bot must be API-based (OpenAI) or built on open-weights model. In other words money-expensive or computational resource-expensive. I wonder if it's possible to engage it enough to cause some harm.

What to write to spin it as much as possible? What's a heavy prompt? Or sequence of prompts. I'm a programmer and I think I could code the Telegram messaging.