FWIW I do try to tell folks what went wrong and what were the perceived gaps through the recruiters. I in fact ask my recruiters to give them known ways to prep

FWIW I do try to tell folks what went wrong and what were the perceived gaps through the recruiters. I in fact ask my recruiters to give them known ways to prepare for our loop. You'd be surprised how many don't even after all this. I am assuming this is because many are just trying to see what's out there, don't have time to prepare etc. But the flip side is we do want candidates who are obviously good AND are excited about the opportunity. All the other concerns here around conflict avoidance / litigation risks are all valid too - That's why I put my feedback through the HR filter. Having said all that, The other angle that always nags me is the fact that our (in all the three companies I've been part of) interview process isn't a perfect measure. In theory, you may be able to crack it if you prepare for it and are well versed in full-stack engineering. In practice, this is just 5 hours of 5 different people trying to look for good signs and feelings and converge after. There just isn't enough information, so most of the time we work towards avoiding a false positive (which have been disastrous in the past). We have tried other mechanics around building a test project etc., but nothing stuck. Those just added a not more non-determinism compared to this model where we know what to expect. I've done more than a 1000 of these, so there are patterns I can fallback on. If I were to start a company, I know how I'd want to hire: It will be just like finding co-founders. But that doesn't scale and probably becomes illegal when it goes beyond the founding members. I do like the doctor group models where everyone joining graduates to a full partner eventually and take a share of yearly profits until they work there: But our current world of software development that's highly leveraged on future performance (stocks), doesn't seem amenable to the same :(. But yes - It would be awesome to see if somehow we can replicate that model.