This is probably the only part I'll give you as far as your comments on this thread, but even this is inane and you're projecting. First of all, the problems with take home tests are the time commitment, which I can understand with the caveat that you do a lengthy pair programming exercise in person. I haven't heard that many complaints about pair programming, but I'm sure I exist so I'll give you that.

I don't think it's the case that people don't like interviews. I think you're making a strawman. I'm not going to go and say that you're part of the cycle of SV engineer being hazed, reproducing it and exhibiting a sunken cost fallacy/bias towards the way you've done it, but it's possible. I don't think this is the ideal way to interview.

What's more, I've been seeing actual improvement on the state of the art in this area. Platforms like Karat et al (not sure if Karat is the best in this area but I did have a good experience with a company that used them and interviewed me well) are actually incentivized to minimize false positives and negatives rather than just one. You might be surprised at how much better a good technical interviewer can be than an average strong engineer.