LLMs work by always predicting the next most likely token and LLM detection works by checking how often the next most likely token was chosen. You can tell the LLM to choose less likely tokens more often (turn up the heat parameter) but you will only get gibberish out if you do. So no, there is not.

If you have vitamin deficiencies, fortified food can also cause you to run at 120%. In some countries (e.g. US) Noodles and flour are fortified with B-vitamins. Maybe that’s what you’re experiencing? It happened to me a few times.

player_hand and dealer_hand are only [DECK_SIZE/2] in length, but in initialize_decks you write zeros into them unti [DECK_SIZE -1]. Since the arrays are located next to each other in memory you end up overwriting the deck array.

And

Looks to me like they’re trying to build something like office 365, but open source. Mostly by wiring other open-source components together I think?

This is probably a better starting point, unfortunately the text is in German: https://gitlab.opencode.de/bmi/souveraener_arbeitsplatz/info