Perhaps they blew it all on "research" and GPU costs?
The current Gemini canvas implementation is pretty wild and can create you an app or website very very easily. Not as easy as ordering a pizza sure, but still even with just a few iterations you have something truly decent. Occasionally you luck out and the "one shot" is precisely what you want, but that typically requires an extensive prompt so you're beyond pizza range then.
Perhaps they were just too ahead of their time in that regard... But then if they'd waited then Google and OpenAi and Anthropic would have just eaten their lunch anyway (provided Claude was not too busy blackmailing you anyway!)
The current Gemini canvas implementation is pretty wild and can create you an app or website very very easily. Not as easy as ordering a pizza sure, but still even with just a few iterations you have something truly decent. Occasionally you luck out and the "one shot" is precisely what you want, but that typically requires an extensive prompt so you're beyond pizza range then.
Perhaps they were just too ahead of their time in that regard... But then if they'd waited then Google and OpenAi and Anthropic would have just eaten their lunch anyway (provided Claude was not too busy blackmailing you anyway!)