Can you share the agent-comparison harness code or point to something similar? I want to learn about benchmarking models in a basic or practical sense.
It depends on the task but I have different Claude commands that have this role, usually I launch them from the same session. The command has the goal of doing an analysis and generating a md file that I can execute with a specific command and the md as parameter. It works quite well. The generated file is a thorough analysis of hundred of lines with specific coded content. It’s more precise that my few line prompt and help Claude stay on rails
I don't see the relation. A "ghost job posting" is not evidence of recent graduates failing to meet the requirements of the past; it is a job posting an employer has no intention of filling to begin with.
> Surely this would be indicated by a glut of unfilled job postings.
I posted a link about Ghost jobs. Then, you said:
>[...]it is a job posting an employer has no intention of filling to begin with.
GP' comment speaks to recent graduates feeling less engaged. Whether it's because they fail to meet the requirements, or the requirements are literally fake doesn't matter. AI isn't used simply to cheat on coursework, but also to erect a de facto glass ceiling viz fake jobs with fake requirements, engagement suffers.
A labor pool's competence drops, especially at the lower/entry-level end, when educational achievement for the labor drops.
To illustrate with reductive absurdity: If every CS student in the market used AI to do all of their coursework, and got a degree still -- that would, among other things, likely reduce appetite on the hiring side.
Right, but there's no way of actually measuring whether or not there's a drop in competency if the earnest job postings aren't there to begin with. I.e. there's no way to test your hypothesis.
