AI Is Robbing Jr. Devs | Hacker News Reader

AI Is Robbing Jr. Devs

benbrougher.tech

16 pointsmakerdiety15 comments

I don’t really get this attitude either way. I expect my junior devs to use all the tools at their disposal, including LLMs, to solve the problems I give them. These problems are rarely so rote that an LLM can just do them- there tends to be a great deal of context building and understating what the customer wants beyond the actual coding. That is a huge part of what I’m really offloading to the junior developer, not just writing the code.

We still discuss the code when delivered and iterate. Devs still receive feed back on their solutions and are held to account if the solutions aren’t of sufficient quality. The social element is still there, we are now just also co learning how to use this new tool.

This results in short term gains of being able to deliver something quickly, but the devs who used LLMs did not learn anything.

The value for a developer is not in simply flipping a piece of code into production, it’s in learning how to find the documentation, how to dig through it to find the functions they need, how to chase down the chain of dependencies for the input and outputs, and all the other incidental exposure they get to related information they might not need for this specific request but may be relevant in the future.

While a company may not care and just wants to ship code, this is talking about what developers lose, not what companies lose. Also a company focused on this type of short term gain is losing the ability to have someone ready to find problems when the code doesn’t work as expected.

I see where the author is going, but I think we need to be realistic that LLMs are taking a cut of where we would normally use a junior dev. the value prop is fairly poor if we expect a junior to leave after 18-24 months and they aren't able to learn enough in a short enough period to get some of the investment back. I'm still all for having junior devs around- I think they provide a LOT of intangible value, but we probably should be realistic about the new world we're heading into.

If you're mid level, you'll be even more in demand as LLM code fixers. But look at industries that didn't build juniors up and see that they have the hardest time recruiting, so much so that PE buys them out and guts everything.

Someone will get that "investment", and statistically, you will get other's "investment" when a former junior joins your team.

We can be realistic, or we can just surrender prematurely. The latter can be dangerous.

We can never capture all the value we create in the world. The more value we create, the smaller proportion of it we can capture.

Education is the quintessential example of this. Preschool is an incredible investment [0] but the payoff happens far away in time in space. While you can't capture the value generated by running a great preschool, you certainly benefit from a community of adults who had strong early childhood education... thanks to investments from other people at some other time & place.

An expansive mindset is required to benefit from these kinds of investments.

[0] https://files.eric.ed.gov/fulltext/EJ1118537.pdf

The problem I see with this is: no-one is born a senior developer. One starts as a junior-level dev, and through trial and error, mistakes and mentoring, grow into an increasingly senior person. You are not just investing into a person who will leave; you are investing in your future senior devs.

Are there any examples of other industries where the entry level was disrupted, so that there was reduced demand for workers below senior level? If that's what happening to software development, could some parallels be drawn?

Apprenticeship has pretty much disappeared, even from trades, but as far as I know, "juniors" have always been in demand, if for nothing else but the lower salaries.

Longshoremen (those "essential workers" who were praised but not paid any better) are in a long negotia-battle against automating some of their jobs away at the moment. They're unionized, so it's the "juniors" who are at the most risk of job loss.

https://apnews.com/article/ports-strike-longshoremen-unions-...

I don’t think what you are thinking of would happen because we’ll just redefine what a “junior” is.

In some industries you see education inflation- apprenticeships may be declining, but you can get a 2 year degree in auto repair at my community college. Similarly, you used to be able to become a lawyer through apprenticeship and self study, so a “junior” in the profession was the apprentice. Lawyers now enter the profession after 3 years of law school, so the bar for what a “junior” lawyer does and is has changed.

More significantly, you’ll see what is expected of a junior developer (and developers in general) change. My understanding is that prior to personal computers and spreadsheets accountants spent an enormous amount of time doing arithmetic. Presumably, junior accountants were expected to do a lot more of the arithmetic than the senior accountants. The invention of the digital spreadsheet eliminated a lot out of this work, but accountancy (and junior accountants) ended up evolving to deliver ever more insights about how a business operates and now there are more accountants than ever.

As a principal software engineer, I have far more problems to solve than I have time or available resources. If a new grad can solve those problems I’ll happily let them do so. There are precious few problems I have that an LLM can largely solve on its own and where that is true the new grad is often better at utilizing an LLM than I am.

I’ll add that these changes place different values on different skills and we are at a transitional moment. I imagine that the junior accountant with incredible arithmetic skills but no interest in computers or saw their career prospects dim with invention of the spreadsheet.

Good points, thank you. I wonder what the requirements and expectations towards newly graduated devs will be like once (or if at all) the industry stabilizes with these new tools. Will they just be expected to stick spaghetti to the wall faster than before, or will the surplus of quantity (of code) lead to an improved appreciation for quality and basics

The present is robbing the future. Old story.

Junior devs can also use LLMs thereby making them better than LLMs.

This is not at all obvious, and I'd argue it's untrue. LLMs are tools that generate plausible sounding text. If that text is correct, they might save you time, but if it's wrong, it's leading you down the wrong path and wasting time.

The less experience you have, the harder it is to tell how bad the answer is, which wastes even more time.

Reply on news.ycombinator.com