I am not exactly a great example ( exposure to work model, ollama, local models play ) and I actually liked gemini upon try in google search ( which is amusingly now banned at work ), but the nice quickly fell into not nice, when it started giving me weird pushback on operation paperclip book ( I am assuming chapter discussing tabun triggered something. This is my only real problem with gemini. By comparison, I am not running into guardrails with gpt nearly as often.

Even “AI mode” isn’t mainstream yet, at least in my observation.

“AI summaries” are, but they seem to be powered by an even weaker model.