There's an awful lot of typos in the little Veo/Sora type clip. "Shar", "Evolue", "Transilate", etc.

Is this all vibe coded? (mdashes ;) ... I cannot see how one can publish agent skills /task capabilities on the site. It seems to be just a list of open resources (most likely also found using claude or ChatGPT ?).

You should probably get a different domain

last night I registered misdirect.ai and discredit.ai ...the general idea is to run agents that prod LLMs into writing the most insidiously difficult to spot incorrect answers to any question, with an army of people to copy/paste those into online forums. The hope is to use existing AI to poison future models, and as a bonus to destroy any remaining credibility people give to anything they read on the internet as a whole. If anyone feels like helping out, there is no money in this, but it feels like a service to humanity. Currently just an idea I had when I was drunk, but, sometimes it takes a band to get you up on Sundays to play music.

Yeah, I was expecting to watch some guy called Claude's COD clips. Instead I get this.

This observation is so true. even the awesome-xxx projects don't really help, or don't reach enough audience, or not the easiest to consume. I hope your project has learned from that lesson.

Do you have any evidence that this works? Can you link the evaluation and result?

Anthropic legal will be reaching out tomorrow if not Monday.

Thank you for sharing

I find your lack of emojis disturbing.