I think the person you're replying to is being a bit alarmist, but in the context of AI, the models available 8 months ago were honestly significantly worse than what we have now.
Simon Willison did a great write up about the advancements of LLMs over the past six months, and it was a pretty cool read.