Remember when GPT-3 came out and everybody collectively freaked the hell out? That's how I've felt watching the reaction to any of the new model releases lately

Remember when GPT-3 came out and everybody collectively freaked the hell out? That's how I've felt watching the reaction to any of the new model releases lately that make any progress. I'm honestly not complaining about the model releases, though. Despite their shortcomings, they are extremely useful. I've found Gemini 3 to be an extremely useful learning aid, so as long as I don't blindly trust its output, and if you're trying to learn, you really ought not do that anyways. (Despite what people and benchmarks say, I've already caught some random hallucinations, it still feels like you're likely to run into hallucinations on a regular basis. Not a huge problem, but, you know.)