Hacker News Reader

Comment by georgeburdell | Hacker News Reader

I don’t understand this logic. If I’m being honest, the most productive people with AI on my team are the two youngest. They don’t have product wisdom, yes, but the gap is so noticeable that I pushed my manager to forego the full-time work experience requirement that we have always had so that new hires don’t even know the old way of working and can just reframe the field in terms of AI.

AI is the biggest threat in over a generation to seniors who aren’t investing in keeping up. Managers without an instinct for AI are an endangered species as well.

View on HN

But the AI transition is different than old technological changes in that you don't really have to learn anything. You don't have to memorize keywords and new concepts like you'd have to for, say, a new JavaScript framework. If you don't know something about AI, you can just ask it and it will tell you.

Basically the only hard part for old people is getting USED to using AI. Some haven't, but it only takes a few weeks to do. After that, experience and taste really comes in handy and that can't be developed overnight.

>If you don't know something about AI, you can just ask it and it will tell you.

If your questions are sharp, probing for detail, and intelligent. Have you seen how people question AIs? It's as if they can't even conceive of the AI being able to supply specific detail.

I'd love to see examples of that! Just from debates on HN alone, I get the idea that there's huge deltas, and I'm getting really really curious.

Point of order: On the one hand I'm not sure people quoting output from software where relevant is entirely what the 'no ai' rule is about. But it's arguable, fine.

For this kind of task, is linking to pastebin etc acceptable instead?

Either way thank you tstrimple for going in and looking!

Yeah that's my feeling about it. I don't code much these days so have not used AI very much in that respect, so for me it's basically a superpowered search agent that can formulate relevant answers based on the details I provide, without the "why do you want to do that" or "tell us what you're really trying to achieve" that you'd so often get on tech forums.

But if I had to use AI for coding I don't think that it would take me very much time at all to get into the swing of it. I wrote code by hand for decades, so I know how to judge whether it's producing something that looks reasonable.

Yep, that's been my experience of it.

Fun note though, my work these days often involves telling AI to solve a problem and me asking it why it wants to do something a certain way.

It often has to patiently explain something like why the bazel rules need to be modified to accomplish my ends goals, and has to reiterate key concepts about a part of the codebase I'm unfamiliar with these days.

Really the only reason I'm there is to make sure it makes the changes as clean and efficient as possible, without unnecessary complexity that's hard to read.

> telling AI to solve a problem and me asking it why it wants to do something a certain way

Just keep your guard up for when the answer to "why" may be just a post-hoc "rationalization".

There are effectively 2 schools of thought on this.

One states that more junior people will excel. Younger people tend to be more hip to new tech trends, more natively immersed in changing times. And the contemporary AI tools eradicate any knowledge & wisdom gap that the seniors may hold over them. Thus, they'll be faster.

The other school states that the most important thing right now is the wisdom of experience. That the AI tools have made everything *but* the wisdom pretty much an afterthought. Thus, there's no point in having someone who doesn't have deep wisdom.

Which is right? Probably a little bit of both, depending on one's exact needs.

From my experience.

There is a reason why youngest developers are more productive. They accept everything Agent suggests. They don't understand or care. They rely on your review of the Agent output from prompt they copy-pasted from JIRA. So you spent your time reviewing it instead of your own work.

From study titled "The Widening Gap: The Benefits and Harms of Generative AI for Novice Programmers" https://dl.acm.org/doi/epdf/10.1145/3632620.3671116

>> Without building this solution, and instead just taking what ChatGPT generated, they had skipped crucial steps in the programming problem solving process, and were now lost

When you skip the problem-solving process and take the LLM's first output, you're not building the cognitive framework needed to debug or maintain that code.

I had a junior happily accept the agent's suggestions, bugs and all. Her productivity ended with her tenure; meat proxies add no value.

The juniors I have seen don't even try to understand anymore what they are building. They don't understand the product, they don't understand the technology. So everything AI is building for them is nonsense. "Reviewing" it is impossible. Some of them think they are being extra clever having Claude process review comments for them. Even if you hire one, this way they are out the door within the month.

I think the problem is evaluation. They can't tell who the good juniors are with respect to the new skills that they are looking for, so they go for experience instead. It doesn't seem sustainable though for obvious reasons.

> AI is the biggest threat in over a generation to seniors who aren’t investing in keeping up. Managers without an instinct for AI are an endangered species as well.

New tech has always been the biggest threat to seniors who aren't investing in keeping up.

Same here. The absolute star in my team is the youngest. But! I think what's happening is that the distribution in this age group has shifted. Essentially the absolute best (high IQ, super ambitious, great communicator, cracked builder types) are now even better positioned than ever before. But the majority of mass in this age category now suffers, i.e. everyone that's "good" or "very good" but not exceptional/brilliant.

Depends on the nature of the work. In highly specialised work the junior would train with a senior while the senior can offload the easier work.

Now the senior can use the AI instead of dealing with a junior.

Plus the biggest danger I’ve seen numerous times is the junior engineer can’t easily tell if the AI is wrong.

They either accept it as correct or spend more time trying to fix AI output with an AI in an area they don’t have the skills for.

In my experience in the past year, new hires, juniors or mid-level programmers (probably even "seniors") who use AI to contribute quickly have a big issue of now ownership or understanding of the code and of the issues that we might encounter. They contribute very little during meeting because they have trouble understanding the subject matter.

We are a network tooling team, and now we think our next hire will be a young network engineer who probably will generate code that we will have to review carefully, but we have to do that anyway, because we can't really afford having our engineers stuck on a basic knowledge of networking issues.

In any case, thanks to AI everything have a SDK so we won't have to make in-house library to automate everything (thousand of hours making libraries for cisco, forti, zscalers and other lost, but to be honest, good riddance), so the demand for engineers in my situation will decrease.

> I don’t understand this logic. If I’m being honest, the most productive people with AI on my team are the two youngest.

It's high variance. A young person can move fast because they lack the accrued assumptions of a senior, but that also means they can quickly move in the wrong direction because they lack the wisdom to identify mistakes before they pile up.

"Use AI" isn't really a skill, getting an implementation is now as easy as google for an answer was for our generation. A senior can trivially learn this. But sniffing out the bullshit and knowing when you're going astray still requires a critical eye that not everybody is going to have.

So it seems reasonable to me that moving more slowly and being more cautious, letting the uncs use their knowledge to keep things from going too far afield, is a really smart idea.

On the contrary, new generations are skipping the fundamentals and using AI instead.

Seniors will become more valuable as that knowledge is lost and AI isn't capable of full replacenent.

I use AI daily and it takes almost no extra brain power to use.

How are the youngest more productive than the seniors? As an aside, I have seen that the newest employees are performing better than those hired during Covid. I think the quality of recent hires are just better because no one is hiring, and so many quality candidates are available.

I didn’t say “most productive”, there was a qualifier attached. A person productive with AI is capable in a different way than the existing folks who are struggling to keep up with their jobs and retool to AI

If you find a proper senior dev - it will be like 10 or more of that junior you described.

Because you can’t shortcut experience. And AI multiplies that.

I’ve seen what ai-focused good senior dev can do and it is scary.

We can’t find seniors like that yet, even though our pay is nearly top tier, because hiring is so broken right now. Most hires come from referrals so it’s recycling the same people between different companies, and those people don’t have good AI experience.

> the most productive people with AI on my team are the two youngest. They don’t have product wisdom, yes, but the gap is so noticeable

Is it Software or another problem domain? What problem are they attacking, what stack/method are they using?

Reply on news.ycombinator.com