Not to put too fine a point on it, but reactions like "you must be suffering from an inferiority complex" or "you need therapy" are exactly the sort of blame shifting one should expect from SV brogrammers. The only voice missing is the one saying "if you can't hack it, quit". The problem is never them, the problem is always you.
I don't see a challenge as evidence that someone is/or believes they are smarter than me. On the other hand I can clearly see that some tangential challenges are given to job candidates "because FAANGs did them" making the potential employer at best a parrot and that they likely feel smarter than those they hire. I don't feel I have an "inferiority complex" but neither am I intimidated by smart candidates. I feel no need to make them sweat or demean them. In fact I would rather hire people who I feel are smarter than me. Wouldn't you?
Programming in a made up language and solving problems like "How many toilets in NYC?" serves no productive hiring goal other than to make the test giver feel smart about knowing the answers while the test taker struggles to to remember if your made up language used let or was it set and why the hell are they asking me about toilets?
Along the same lines, I feel if someone is asking typical programming challenges like "balance a tree" then they are wasting their time and the candidate's time. That is an implementation detail that you can lookup. Likely there is a library that will do it for you.
In my opinion a clearly superior question is "Why might you want to rebalance a binary tree?". A better version of the afore mentioned task would be "Now you have explained why one might want to balance a tree, using any and all available resources show me how you might do that in practice." With top marks going to the candidate that finds the standard package and calls balanceBST().
I hire and manage an international staff of coders from North America, Eastern Europe, South America, and India and frankly I could care less if someone can implement a recursive tree walker or regurgitate a k-means clustering from memory. There are literally libraries maintained by others to do that. If for some reason I cannot fathom, we needed our own, I would damn well expect my staff to use resources like google, wikipedia, and open source to assist in writing it rather than have us all thank our lucky stars we work at that company where everyone knows how to hand roll a tree balancer only to find that we wasted a sprint dealing with an "off by one error". If one feels they absolutely must test for understanding of basic concepts such as recursion, composition, or inheritance, there are plenty of established interpreted languages to use.
Do you (rhetorically speaking) really want to make your job applicants uncomfortable by shoving a made up language at them that only you and your buddies know? Is your stack so mutable that you need candidates that can deal with regular resets to the platform of the day? Hopefully great candidates realize that a test in a homebrew language is a dick move and likely a signal of a larger organizational problem and politely thank you for your time then take their expertise elsewhere. There are plenty of other opportunities.