I'll be the victim.

This idiot Karl is messing the internet all up with his derogatory speech. Let's make the world a better place by filtering everything he says so that it's nice!

Luckily I am Karl, so I will be able to figure out what the heck is going on in my life, by being the person causing the problem. I really do have a problem with posting crazy things to mailing lists on the internet, so it would be pretty nice if some completely transparent system that I am in complete control of, would mutate anything rude or harmful I might say to be nice, so that I don't make enemies during psychotic breaks.

Deploying a mitm attack on the stupid idiot typing this email is a challenge. There are a few difficult parts:

- writing code to mutate packets live without disrupting network sessions
- figuring out and implementing a workaround to SSL if the guy uses it

- training a language model to mutate his speech in the desired manner

This idea seems _so incredibly cool_ to me, that I have actually already started step #3 in a tiny, microscopic, nonexistent manner.

https://vast.ai/ is a service that provides brief access to high-end machines for training transformer models to make deepfakes and such.