What's disappointing about agent-dev interviews
is most questions are behavioral and cognitive
Favoring candidates who can stun interviewers with talk
This mechanism doesn't filter for people with taste
E.g. how do you evaluate harness engineering optimizations?
Maybe design a benchmark system, accuracy up xx, errors down xx—scripted talk.
I hate that framing
Good engineering results speak for themselves
I don't need to self-prove
It's confidence in craft and taste
I chose this approach
I shipped the effect
I used it myself and felt improvement
Qualitative perception is enough
Why anchor value in numbers
Form over meaning
Just for easy performance reports?
Same old org-collaboration ghosts
So if a fresh grad
has a silver tongue
plus a bit of hands-on experience to structure and deepen answers
but little engineering taste or depth
and gets the same high pay as my level
I hate seeing that
A distorted industry
How to break out?
Let others see my taste?
Still feels like starting from real-world problems
Design a system that solves one
The process would show it
But if it's something others have done
or made for interviews / eyeballs
I won't do it
Skip to content
What's disappointing about agent-dev interviews
From X · @therealfreddyx
Thoughts
mew～
What's disappointing about agent-dev interviews
is most questions are behavioral and cognitive
Favoring candidates who can stun interviewers with talk
This mechanism doesn't filter for people with taste
E.g. how do you evaluate harness engineering optimizations?
Maybe design a benchmark system, accuracy up xx, errors down xx—scripted talk.
I hate that framing
Good engineering results speak for themselves
I don't need to self-prove
It's confidence in craft and taste
I chose this approach
I shipped the effect
I used it myself and felt improvement
Qualitative perception is enough
Why anchor value in numbers
Form over meaning
Just for easy performance reports?
Same old org-collaboration ghosts
So if a fresh grad
has a silver tongue
plus a bit of hands-on experience to structure and deepen answers
but little engineering taste or depth
and gets the same high pay as my level
I hate seeing that
A distorted industry
How to break out?
Let others see my taste?
Still feels like starting from real-world problems
Design a system that solves one
The process would show it
But if it's something others have done
or made for interviews / eyeballs
I won't do it
is most questions are behavioral and cognitive
Favoring candidates who can stun interviewers with talk
This mechanism doesn't filter for people with taste
E.g. how do you evaluate harness engineering optimizations?
Maybe design a benchmark system, accuracy up xx, errors down xx—scripted talk.
I hate that framing
Good engineering results speak for themselves
I don't need to self-prove
It's confidence in craft and taste
I chose this approach
I shipped the effect
I used it myself and felt improvement
Qualitative perception is enough
Why anchor value in numbers
Form over meaning
Just for easy performance reports?
Same old org-collaboration ghosts
So if a fresh grad
has a silver tongue
plus a bit of hands-on experience to structure and deepen answers
but little engineering taste or depth
and gets the same high pay as my level
I hate seeing that
A distorted industry
How to break out?
Let others see my taste?
Still feels like starting from real-world problems
Design a system that solves one
The process would show it
But if it's something others have done
or made for interviews / eyeballs
I won't do it