OK so this is what DeepSeek came up with as a solution to "The need for a more sophisticated compression algorithm." (And I'm quoting myself here)
Hym "The article said that it uses 'visual something' or 'visual perception' and fit 7 to 20 times as many tokens. I don't know. The article didn't explain it very well and then the news started and then I went back to find the article and it had left my feed... But good job! Brilliant stuff guys. I'd love to see a more in-depth explanation of it."
by Hym Iam October 22, 2025
Get the Visual Perceptionmug.