But not necessarily in a good way. This introduces a bias into the users prompt which is not there without the suggestion. So, arguably, this feature makes the users inputs less useful for training because they are partially impacted by model output.
This is similar to blog articles since 2023 or so onwards being less useful for model training since more and more of them are based on LLM output to varying degrees.
Good point, if we thought the ensloppification was bad now, imagine how bad it will be when the AI is slop prompting itself.
i believe they sometimes show no suggestion at all, fwiw.
It’s been a long time since I’ve seen no suggestion at all. If it doesn’t have anything else to suggest, it will suggest committing.