I would personally feel way better about interviewing for jobs if it was like peer reviews during a performance review cycle.

I've written a bunch of words in this thread about how I'm skeptical of your second paragraph. I recognize that you and many people I know and have discussed this with are adamant about this being true, and yet I remain unconvinced. I have interviewed many people, some of whom attempted to talk up their previous experience unconvincingly, and some of whom were impressive but nevertheless unsuccessful after being hired, for various reasons. I still haven't come across this person who can go into depth about their previous projects, while being a total incompetent bullshitter. I'm certain such people exist. I'm unconvinced that they are as pervasive as all the discourse on this would have me believe.

But honestly, does anyone here have experience with doing this kind of interviewing at scale, for experienced software engineering roles? I've been in the industry for coming up on a couple decades, and I have been involved in doing lots of interviews at times, and in that whole period of time, at every company where I've worked, we did the standard multi-round whiteboard / coderpad interviews. Do other folks here actually have recent experience "hiring at scale" in this industry, with a process focused on candidates' experience? Who is doing that?

And also, the question is not "are there ever false positives?" or even "is this biased toward a certain kind of false positive?". Nobody thinks any way of hiring is perfect. Even ignoring bullshit artists, sometimes very competent people just aren't a good fit for the actual demands of a particular job. The question is one of tradeoffs. Are the failure modes and biases of a particular process worse than those of another one?

To me, the current standard process comes at an enormous cost. At any job I've ever had, every time I think that maybe my time there has run its course, I immediately think, ugh, but I'll have to go through the f**ing interview process. I'm not a person who does research on this, so I don't have data or anything, but I must not be alone in this, and I think it is likely a meaningful friction in the job market. (Which, I guess is good for employers, so it probably makes sense that they like the status quo!)

I'm fully aware that according to the internet, there is an epidemic of bullshit artists who can go deep on the architecture and tradeoffs and their contribution to the things they worked on, without having actually contributed to those things, but I dunno, the narrative just doesn't jibe with my anecdotal experience.

My only uncertainty here is that I do think I have been very fortunate in the people I have worked with in my career, so I might just be getting lucky. But I have truly never worked with someone who is hired largely on the basis of a strong resume, but genuinely can't grok fizzbuzz, or whatever the contemporary equivalent of that is.

I also recognize that you ran a real company doing real work on this, so I'm generally inclined to defer to your wisdom...

You can tell that I'm very torn on this, because the conventional wisdom is so strongly against my perspective on it, and I generally put a good deal of weight on conventional wisdom. But man, I dunno, it all really feels like an inertia thing that I increasingly question the original foundations of. Sometimes the emperor actually doesn't have any clothes on...

But then I've also read a huge amount about interviewing "correctly" over the years, probably starting with the fizzbuzz article, and eventually participating in "Big Tech" interview panel training, etc. And all of this directly contradicts this natural intuition that I have, and which your comment is espousing.

So I'm honestly left with pretty strong cognitive dissonance about it at this point. Are we wrong? Or is everyone else wrong? How can this consensus on the "right way" have become so ingrained for so long while being so wrong?

(I also haven't been involved in a lot of hiring or at a big tech company since 2022, so I also have no idea how things have evolved to adapt to the advent of AI tools. Surely nobody is doing the same kinds of whiteboard problems as they used to do! Right?!)

This is a counterexample to your argument, not evidence for your claim. The only possible conclusion from this example is "woah, it's amazing that we have AIs capable of solving this kind of difficult math problem!", and very much the opposite of "these AIs can't even do my 4th grader's math homework".

I would be shocked if I was unable to solve 4th grade math homework with any of the contemporary frontier models. I spend most days using them to do significantly more complex things than that.

One of the reasons I'm increasingly skeptical of this prediction is that I've now lived past a few of the dates I heard people put on the achievement of this level of superintelligence in previous years.