In a world where Meta is releasing actual code and model weights to do amazing things, why does Google/Deepmind still think anyone cares about these kind of fak

In a world where Meta is releasing actual code and model weights to do amazing things, why does Google/Deepmind still think anyone cares about these kind of fake puff piece blog posts that never lead to anything that anyone can ever use? I mean, at least let people submit inputs and generate outputs if you're not going to release the actual model. When Google does finally release something that people can use, like Gemini, it's absurdly bad compared to OpenAI, Anthropic, Meta, Mistral, etc. despite all the fake benchmarks. I guess I sound super negative, but at this point I treat their announcements about the same as IBM's-- just a bunch of vapor nonsense. It didn't help when their material science paper turned out to be totally fake, either.