This is very well explained! Influencers convincing students not to study CS and companies no longer hiring juniors have huge long-term implications for the field.
Tbh it’s heartbreaking to see it happening when it’s such a great time to get into computer science.
While we aren't hiring like we were in 2021-22, we're still hiring a nontrivial number of new grads through student programs. But there's a huge cliff between the top of the big tech mountain where I work for the day job and everywhere else. Hiring managers I know—and myself included—are being much more "picky" about hiring. I know some current computer science students whose summer internships have already been canceled. 😞
I the jury is very much out on this topic. I’ve gone back and forth on this a few times and this is a hot take.
Anyone saying that a junior engineer is capable of editing code created by AI can’t possibly be considered credible. The whole shift to senior engineers is based on their ability to leverage AI in practice. Junior engineers lack the experience necessary to do this safely. If juniors were capable of editing AI code the shift would be going the other way!
Software engineering isn’t done in layers. Junior engineers don’t just write boilerplate and pass that off to a senior to complete or enhance. Seniors write as much boilerplate as anyone else, they just do it with more confidence and experience. They are just far better at determining the quality of the code and design. The way the AI tools ship today they are there ready for you to fuck yourself with. If I get lazy for a single prompt I can waste an entire day on sunk cost.
For most companies the internal talent pipeline doesn’t matter much. If you’re employing thousands of engineers, sure if you stop hiring juniors your pipeline may eventually dry up. My team is 100 engineers. We pay everyone at the same level the same so beyond the knowledge of our company and systems, which is absolutely real but not necessarily always a plus, especially from the standpoint of a junior, there is little incentive anymore to hire juniors. If you’re running a very large organization you’re much more likely to have the resources to make AI safe to use so hiring fewer juniors will not actually hurt you, except in the aggregate if everyone does this. In other words if you have the quality box defined well-enough around the code you can throw away the implementation every single day. AI could rewrite your entire app continuously.
I don’t have the luxury of timelines this long, nor do I employ enough engineers to make an impact overall. Plus, the tools are getting better. Yes they are absolutely terrible now but I am finally seeing the tool makers actually addressing repeatable workflows, guaranteeing that rules are sent on every request, etc. If I was able to get the results I get from Windsurf and Cursor for a week on a greenfield project for a month or a year I won’t have to hire any engineers until everyone on my team retires.
I am still hiring junior engineers but fewer than before. Everything is a trade off and hedging is important. In order for any organization to weather the uncertainty ahead we must build a bridge to the future while we drive on it.
If you treat juniors as task bots, they will be replaced by real task bots.
The career ladder in traditional organizations is badly broken. Seniors give the crappy tasks to juniors while they insist on overseeing the entire job. This needs to change. Instead of giving juniors concrete, small tasks, they should get the entire job. But only one that has a small risk profile. Don't let junior accountants do bookkeeping. AI can do that. Instead, give those juniors the smallest clients.
You certainly touched on a lot of points! There is a major aspect to junior level devs heavily reliant on AI tools that I don’t think was addressed, at least directly: AI tooling is requiring a lot more active review from seniors and more often.
You do talk about how mentorship has to evolve, which is very true. But I am experiencing the need for a lot more mentorship than in the past. AI tooling is an accelerator and verbose, meaning that the need to review and retool code has increased in frequency and increased in time as juniors are able to move faster and produce more. I’ve experienced it to be an issue because it is taking a lot more time from seniors than the past.
It is a problem, and there will need to b a solution and an evolution, but I haven’t found the perfect approach yet.
While I absolutely concur, the majority of the vibe coding noise is often coming from advocates, startups, and other short term thinkers. These people aren't investing in the future nor thinking beyond the exit ramp. Until we fall into the trough of disappointment for this hype cycle, those people are still going to generate a plethora of bad ideas and hallucinations. 😉
Your writings are so encouraging and thorough. Thank you! This gives me more clarity + hope into what my career trajectory could look like.
Thank you for the very kind words, Ivy!
This is very well explained! Influencers convincing students not to study CS and companies no longer hiring juniors have huge long-term implications for the field.
Tbh it’s heartbreaking to see it happening when it’s such a great time to get into computer science.
While we aren't hiring like we were in 2021-22, we're still hiring a nontrivial number of new grads through student programs. But there's a huge cliff between the top of the big tech mountain where I work for the day job and everywhere else. Hiring managers I know—and myself included—are being much more "picky" about hiring. I know some current computer science students whose summer internships have already been canceled. 😞
I’ve found jobs to be available but highly targeted. Many roles are looking for devs with 5+ years specifically in AI.
Yup. It's no difference than people who were looking for ten years of Kubernetes experience when the tech was a nascent three years old. 🙄
This is a brilliant and in-depth article! I'm not just saying that because my comments are in it :D
I the jury is very much out on this topic. I’ve gone back and forth on this a few times and this is a hot take.
Anyone saying that a junior engineer is capable of editing code created by AI can’t possibly be considered credible. The whole shift to senior engineers is based on their ability to leverage AI in practice. Junior engineers lack the experience necessary to do this safely. If juniors were capable of editing AI code the shift would be going the other way!
Software engineering isn’t done in layers. Junior engineers don’t just write boilerplate and pass that off to a senior to complete or enhance. Seniors write as much boilerplate as anyone else, they just do it with more confidence and experience. They are just far better at determining the quality of the code and design. The way the AI tools ship today they are there ready for you to fuck yourself with. If I get lazy for a single prompt I can waste an entire day on sunk cost.
For most companies the internal talent pipeline doesn’t matter much. If you’re employing thousands of engineers, sure if you stop hiring juniors your pipeline may eventually dry up. My team is 100 engineers. We pay everyone at the same level the same so beyond the knowledge of our company and systems, which is absolutely real but not necessarily always a plus, especially from the standpoint of a junior, there is little incentive anymore to hire juniors. If you’re running a very large organization you’re much more likely to have the resources to make AI safe to use so hiring fewer juniors will not actually hurt you, except in the aggregate if everyone does this. In other words if you have the quality box defined well-enough around the code you can throw away the implementation every single day. AI could rewrite your entire app continuously.
I don’t have the luxury of timelines this long, nor do I employ enough engineers to make an impact overall. Plus, the tools are getting better. Yes they are absolutely terrible now but I am finally seeing the tool makers actually addressing repeatable workflows, guaranteeing that rules are sent on every request, etc. If I was able to get the results I get from Windsurf and Cursor for a week on a greenfield project for a month or a year I won’t have to hire any engineers until everyone on my team retires.
I am still hiring junior engineers but fewer than before. Everything is a trade off and hedging is important. In order for any organization to weather the uncertainty ahead we must build a bridge to the future while we drive on it.
If you treat juniors as task bots, they will be replaced by real task bots.
The career ladder in traditional organizations is badly broken. Seniors give the crappy tasks to juniors while they insist on overseeing the entire job. This needs to change. Instead of giving juniors concrete, small tasks, they should get the entire job. But only one that has a small risk profile. Don't let junior accountants do bookkeeping. AI can do that. Instead, give those juniors the smallest clients.
I wrote about it here: https://substack.jurgenappelo.com/p/ai-wrecks-the-corporate-career-ladder
You certainly touched on a lot of points! There is a major aspect to junior level devs heavily reliant on AI tools that I don’t think was addressed, at least directly: AI tooling is requiring a lot more active review from seniors and more often.
You do talk about how mentorship has to evolve, which is very true. But I am experiencing the need for a lot more mentorship than in the past. AI tooling is an accelerator and verbose, meaning that the need to review and retool code has increased in frequency and increased in time as juniors are able to move faster and produce more. I’ve experienced it to be an issue because it is taking a lot more time from seniors than the past.
It is a problem, and there will need to b a solution and an evolution, but I haven’t found the perfect approach yet.
While I absolutely concur, the majority of the vibe coding noise is often coming from advocates, startups, and other short term thinkers. These people aren't investing in the future nor thinking beyond the exit ramp. Until we fall into the trough of disappointment for this hype cycle, those people are still going to generate a plethora of bad ideas and hallucinations. 😉