A 49 years old java programmer even complained that the test isn’t fair
And you had to mention their age ... why, exactly?
As opposed to "a left-handed Java programmer", or "a Java guy wearing a Slayer T-shirt"? No -- for some reason, with this candidate -- their age was especially relevant to you.
Why?
But then people try to overdo it and deal with UTF-8 before having the basic algo working.
Because of one too many experiences with live coding interviews, (on appearances) pretty much exactly like yours -- in which one gets flushed (or otherwise dinged) for failing to proactively treat such edge-case details as encoding gotchas, zettabyte-scale strings sharded across multiple time zones, etc.
I’m pulling my hair wondering what’s this parallel world I’m in, or whether I’m introducing a bias which makes everyone fail (2 interviewees per month for 5 months).
In fact, you are introducing just such a bias. There has recently been good, solid research done into the fact that performative (that is: on a whiteboard, in front of a group) "programming" tests are really basically high-pass anxiety filters. That is, implicitly designed not to measure programming ability, but the (studied) ability to ... successfully complete peformative rituals in front of random strangers you've never met.
If you aren't aware of this research (or your capacities for empathy aren't at a level where this is intuitively obvious to you) -- then you shouldn't be conducting these kinds of interviews in the first place. No matter how good you are at programming per se.
And you had to mention their age ... why, exactly?
As opposed to "a left-handed Java programmer", or "a Java guy wearing a Slayer T-shirt"? No -- for some reason, with this candidate -- their age was especially relevant to you.
Why?
But then people try to overdo it and deal with UTF-8 before having the basic algo working.
Because of one too many experiences with live coding interviews, (on appearances) pretty much exactly like yours -- in which one gets flushed (or otherwise dinged) for failing to proactively treat such edge-case details as encoding gotchas, zettabyte-scale strings sharded across multiple time zones, etc.
I’m pulling my hair wondering what’s this parallel world I’m in, or whether I’m introducing a bias which makes everyone fail (2 interviewees per month for 5 months).
In fact, you are introducing just such a bias. There has recently been good, solid research done into the fact that performative (that is: on a whiteboard, in front of a group) "programming" tests are really basically high-pass anxiety filters. That is, implicitly designed not to measure programming ability, but the (studied) ability to ... successfully complete peformative rituals in front of random strangers you've never met.
If you aren't aware of this research (or your capacities for empathy aren't at a level where this is intuitively obvious to you) -- then you shouldn't be conducting these kinds of interviews in the first place. No matter how good you are at programming per se.