Big mistakes = Doing something that is actively harmful or useless by their own lights and values, i.e. doesn't help them achieve their life goals. (Not: Doing something that isn't in line with your values and goals.)
A lot of people think that others in the EA-ish community are trying to do something impactful but end up doing something harmful or useless. Sometimes they also work on something that they are just not very good at or make other big mistakes. A lot of people never end up telling the other person that they think they are making big mistakes. Sometimes people also just have one particular argument for why the other might do harmful or useless work but not be sure whether it's a bad overall. This also often goes unsaid.
I think that's understandable and also bad or at least very costly.
Epistemic status: Speculation/rant. I know of another person who might post something in this topic that is much more rigorous and has actual background research.
That said, I often don't do a great job at this myself and think telling others you think their efforts would be better spent elsewhere also has significant costs, both personal and on a community level.
So, where does this leave us?
Finally, you don't have to do any of this. Sharing your honest thoughts can cost you a lot and you should spend your resources on whatever you want to spend them on.
You can give me anonymous feedback. I really care about what I do with my life - If you think I could be doing better, please tell me. I'm also happy to receive short, low quality, low information feedback, although the opposite is preferred of course. If you wanna be extra fantastic and give me non-anonymous feedback, so I can engage with you, I will be over the moon and very grateful - but don't let the perfect be the enemy of the good.
Also, if you think you have important thoughts to share with someone but you don't know how, feel free to dm me. I wrote this post in one go, so maybe I'll have more useful tips for how to actually go about sharing your honest thoughts in individual cases :) I might take a while to respond or decide not to engage in detail though.
[unsure whether to up- or down-vote. This seems too high-level and over-general on a topic where details and nuance can easily override far-mode preferences. But it's important and feedback on errors is one of the best ways to improve one's rationality (on any dimension of error - modeling, weighting, implementation). ]
You imply in your closing (which I like a lot - it's a concrete request and offer) that this feedback should be private. I wish you'd mentioned that sooner, and acknowledged that it may be happening more than you think, since you won't see most of the feedback to others.
I also think that's a little at odds with the type of feedback that should be given to larger projects or ambitious people trying to have a large impact and failing. This feedback should be BOTH public and private. The ways they're doing harm should be available to the people being impacted, not just to the actors. The distinction between individual decisions and organizational or group (small or large) decisions is very important here too - there are different methods of giving different dimensions of feedback, and being good at one is almost unrelated to the other.
Finally, I wonder what purpose you had in specifying "your friends" in the title. I do a lot of 1:1 discussions and performance reviews with software engineers - giving and getting feedback on work topics is a big part of how I improve, and how I improve my teams. I try to give (and seek) advice and evaluation in most aspects of life, but "my friends" is probably the group where I'm least focused on improvement through this channel (meta-discussion about how to improve), and more focused on fun and improvement by object-level trial-and-error (new activities, games, topics of conversation).