Agreed. Too little and too late.

Is he not right though? Have you not interacted with a once star senior engineer who now operates at the level of a low-quality junior because they refuse to use AI?

There was a point a few decades ago with architecture where people moved off blueprints to CAD files, and everyone had to learn CAD or be pushed out, which lead to terrible churn and people I know got kicked out of architecture for good. This is a similar situation; they're not kicking out engineers for being 5% less productive, they're re-interviewing to find engineers who are 1/10th or 1/100th as productive as their peers. It sucks, and I feel bad for people who have to fear for their job, but it's a necessary action

I have yet to see data that supports this claims. In the contrary all I have read and observed point in the opposite direction. Also I have yet to see an easy way to assess the performance of engineers. Lines of code it is certainly not!

This doesn't even make sense. You don't go from star senior to low quality junior by just doing what you've been doing and refusing to prompt the slop machine. If anything you're retaining skills that everyone else is losing. Prompting an llm is not a skill, the only thing this "re-interview" assesses is if you're using the magic box like the idiot CEO wants you to. Which is another way to say, are you de-skilling yourself fast enough, because that's what the CEO thinks is best for the company right now.

<< Have you not interacted with a once star senior engineer who now operates at the level of a low-quality junior because they refuse to use AI?

I can honestly say no. I might question your premise a little. Lets hypothetically say you got Karpathy level senior; clearly a star and beam of sunshine in an otherwise dreary world. Is he now less of a star just because he won't use AI?

It is a real question, because if he is then the question becomes why.

Did he simply stop working as much? Or is he working the same amount, but it looks like he is doing less? If it merely looks like it, we may well have office situation on our hands, where rank and file may end up turning to management and asking themselves and the management one simple question:

What exactly it is you are doing here?

I know for a fact good management can enable some things, but I am starting to question the blind fervor I seem to see more of lately. If you can't motivate Karpathy or worse, you now are suddenly are unable to see his value, then maybe it is not Karpathy's fault.

<< This is a similar situation; they're not kicking out engineers for being 5% less productive, they're re-interviewing to find engineers who are 1/10th or 1/100th as productive as their peers

And this may be the big part: productivity. The problem is that productivity along the lines of silly metrics like: lines of code, agents commits per engineer are.. uhh.. not useless exactly, but, how do I phrase it a sensible way so that it reaches at least one stray manager that watches this conversation,

*really fucking dumb metric to treat as a measure stick.*

But who am I to judge.

<< but it's a necessary action

None of it is necessary. By definition it is all elective. That said, the nature of that relationship goes both ways and, I will be super subtle here, right now it is not the engineers, who can't do it all.

Quite the opposite, management could find itself in a pickle now, because it has neither skill to drive the development, practical knowledge that you only get from getting to close to the abyss, 'taste', theoretical foundation, curiosity or even basic exposure that would it let develop either of the previous facets.

Make no mistake, a lot of us choose to do deal with some of the idiocy for a stability gain. But, if there is none, why do I need you? I might as well try to strike out on my own and watch the AI coded towers fall ( being mildly dramatic for visual purposes; amusingly, I don't think I am that far off ) .

> Is he not right though? Have you not interacted with a once star senior engineer who now operates at the level of a low-quality junior because they refuse to use AI?

I don't even know where to start. Let me just say it that by using LLMs you voluntarily decide to let the atrophy happen to your skills, knowledge and experience. You do things without understanding them, or with vague understanding only. Your thinking process becomes completely dependent on something that you find more and more difficult to verify (if you even care). Why would anyone do such a thing to oneself?

No, we are not there yet. Right now, AI is de-skilling the engineers who are using it.

Sure, they may churn out a lot more code, but what is the quality of that code? What is the security of that code?

Can you prove that the code they are generating is actually having a material impact on the bottom line?

If the AI tools suddenly became too expensive to use, could those engineers go back to doing things the way they once did?

If you want to vibe code, feel free to do that on your own personal projects. For everything else, you need to be able to demonstrate that what you’re doing is not just fluffing your lines-of-code metrics, or your tokenmaxxing metrics, and actually demonstrate real-world positive impact on your bottom line.

Maybe in a few years, we will get to a point where no one will need those skills anymore, and all that stuff will be done locally, so we don’t need any more massive datacenter projects that suck down all the power and all the cooling and create more problems than they solve.

But we are not there today. And the way that the AI bubble is already crumbling, I’m not convinced that we will get there any time soon.

AI is maybe a 10% productivity increase for the entire job, even for the most enthusiastic users.

It’s useful, but it really doesn’t matter that much.

AI is maybe a 10% productivity increase for the entire job _if you maintain previous levels of quality_.

I see AI's main value-add right now as a ready-made excuse/bludgeon to finally let CEOs explore exactly how dogshit they can make everything while still having users.

Why can't they just say something like, "I really shouldn't have said X Y Z. It was insulting and I'm embarrassed that I chose that way of speaking to an employee. As a leader, it's important that I be encouraging and receptive." In other words, don't vaguely allude to what happened by saying there was "a communication failure." CEOs like to act like their subordinates are children and can't tell the difference.

Because, in the CEO mindset, admitting fault is a sign of weakness and poor decision making. In a position where the only value you provide is trust and good decisions, that's a death sentence.