So, the interview can now be 2 leetcode hards in 20 min. Earlier, it was typing solution code from rote memory. Now it is furious vibe-coding copy-pasta from one browser window to another.
More seriously, what will the new questions look like ? In the age of LLMs how does one measure ability objectively. Is it with extreme underspecification of the probem ?
The value of an employee who says “I don’t know how to do that” or “I’ll need to ask my coworkers for help” versus one that says “I am sure I can figure out just about anything by googling” is night and day, and I think the same is true with AI.
Half of the battle is going to be knowing what to ask for.
Lastly I’d like to point out that it makes general sense to test people on the real tools they’ll be using to get their work done. E.g., you wouldn’t get much value testing a bus driver on using a manual transmission if all your buses are automatic. Most corporate leaders are expecting and even demanding that their employees use AI tools to get their work done.
> Most corporate leaders are expecting and even demanding that their employees use AI tools to get their work done.
Imagine for a second that I am an aspiring monopolist. Wouldn't a great way to achieve this be to make people believe that I am their agent, when I am really their adversary? Would I really offer them in good faith a tool that they could use to compete against me? Or would I try to get them to accept a trojan horse that destroys their profits from within. Once, I have siphoned enough profit to significantly damage the business, I can come in, borrow money to buy the company that I basically destroyed and then sell off the parts to an ignorant buyer who doesn't realize who badly I have wounded this once good business, or I just write off the loss for tax purposes.