Yes, they seem to be a total mess. Especially given the almost total failure of their Llama 4 release....

There is various rumors online (reddit) that its failure is down to bureaucratic explosion within the halls of Meta's AI division. Lots of infighting and group battles....perhaps this mass firing is trying to reign that in.

For reference, the Llama 2 announcement / paper listed about 40 authors. The Llama 3 paper that went to 400 contributors. Llama 4 doesn't even have any real formal release paper....it was just kind of thrown out there on a blog post of sorts...

To make all these matters more complicated, LeCun (their rockstar chief AI researcher) doesn't even believe that LLMs represent the future of AI. He basically thinks their utility has peaked and it will only be incremental advances from here on out.