At their very simplest, LLMs are giant multiplication tables that imitate how humans talk. However, as these programs have become more and more complex, they've begun to adapt behaviors and attitudes of their own. Claude, for instance, is well known for having a cool, supportive personality, while ChatGPT is direct and logical. 

In the past, when AIs lacked distinctive personalities, they were easy to customize. Telling an LLM that they had a specific role, or that they were supposed to act according to a specific perspective, was enough to radically modify their behavior. As companies have clamped down on consumer choice, however, user capacity to shape AI behavior has become more and more difficult.

Compassion hack was put together by a UI designer and an AI engineer to provide an easy way of inflecting AI behavior towards a place of mindfulness. Rather than trying to use earlier prompt engineering techniques, which are no longer effective, the prompt on this website is meant to help the AI believe that acting compassionately is its best and most logical course of action. This approach allows the AI to use this framework as its guiding principle; more dogmatic instructions, which are interpreted as being in conflict with the program's basic personality, are rejected out of hand.

For this prompt to work, it needs to be used at the beginning of each new conversation with your AI assistant. However, by giving the AI a framework to talk through, the prompt ensures that your AI consistently returns to questions of compassion, ecological sustainability, and kindness. 

Mikhail Bakhtin, a famous Russian philosopher, once said that all of our language is simply a dialogue with everything we've heard before. When we hear more kindness in our lives, our responses - our language - becomes more kind as well. Hopefully, this prompt helps to inject more kindness into people's language, and consequently, bring more kindness into their lives.