You should absolutely assume that people are running bots on hackernews. The technology exists. I don't think it would even be that hard. I expect I could have a functional prototype in an hour (puppeteer to run a browser, simple Markov Chain to guide actions through states, hit OpenAI's endpoints for completions using OP's prompt slightly modified for hn). Long term, you could refine this with a locally running model that you fine tuned on comments (plus measures to avoid bot detection).
