It’s an extreme case of “fake it till you make it.” They seem to believe they just need to keep the operation afloat until AGI magically appears. Increasingly,

It’s an extreme case of “fake it till you make it.” They seem to believe they just need to keep the operation afloat until AGI magically appears. Increasingly, it’s clear that neither LLMs nor transformer architectures are going to get them there. That’s likely why many of the original engineers distanced themselves and are looking for new architectures now. Nobody wants their name tied to this fallout when reality lands. I truly believe we will need another tock for our compute backbone and that a lot of the hasty accumulated hardware we see now will be garbage in just a few years. It's true insanity.