@binsk rofl good point. Idk, I'm of the mindset of not contributing to empty gesture culture. I just think the model could do a better job, if needed at all. There are plenty of ways to soften a blunt statement without changing its meaning!

Like by adding detail, or an emoji. Or, in that exact example of "this is never going to work" instead of stating something you believe as fact, stating it as a prediction which is actually a more true statement anyway (unless you actually literally know the future in this scenario based on scientific evidence of what the outcome should be, like adding 2 and 2, knowing nobody is adding anything else behind your back, and expecting 5.). So like "I think this won't work" or even "I predict this won't work".
If that's not enough, the model can stack on the adding detail method. "I think this won't work, but I trust you to do the right thing" or "but I'm willing to wait and see" would be solid recommendations, which obv the user can then go and confirm if true or not and pick their fave.