AI disrupting the path to seniority
Posted by sliceohpizza@reddit | ExperiencedDevs | View on Reddit | 48 comments
I'm neither an AI doomer or boomer, but the traditional software engineering career development model where junior developers build expertise through scaffolded, routine coding tasks is being disrupted by AI.
As an engineering leader I'm starting to think through how we hire and develop junior talent so we don't hollow out our senior engineer pipeline.
What's everyone's thoughts on:
What competencies should we help junior developers build instead?
How should teams restructure their work and mentorship to create a sustainable path to seniority?
SabretoothSnapdragon@reddit
Being a senior engineer is not really about having superior coding skills (which AI can help with), but more about helping people on a team work well together, understanding architectural patterns and knowing which ones can best meet business needs, unblocking other team members - all things which AI does not really change.
So with that in mind, I don't really understand the premise of your post. Building these skills is more or less the same, regardless of whether team members are using AI assisted coding.
sliceohpizza@reddit (OP)
I agree with you fully that the competencies which define a good senior engineer often have little to do with coding.
I disagree with you that building these skills will be the same, and that was the premise of my post. For example, I'm already seeing "review fatigue" in my teams as the amount of code to review increases with AIs ability to generate. So one question I'm asking myself is how structurally we ensure Seniors have the mental bandwidth to provide that feedback.
Groove-Theory@reddit
Yes the skills are the same, but the PATH to getting there (according to OP) is different. Quietly the "junior takes bug and gets hand-held how to walk it" to "mid level with more confidence can do something with explicit instructions is changing if you can just prompt whatever you want.
All of that has historically built the senior who asks "why SHOULD we build this" or Staff levels of more meta-level impact.
If that traditional path is gone (I'm not convinced either way) then it makes sense to note of some sort of other methodology to train our most green engineers to those same skills needed. Else without any guidance, how can we expect our freshest (and most vulnerable/malleable) engineers to achieve the same skills that we have?
It's definitely a fair question.
psyyduck@reddit
OP is a bot.
sliceohpizza@reddit (OP)
Ok boss. On what basis are you making that determination?
t-tekin@reddit
To be honest, market is filled with such strong senior engineers at the moment, I don’t see this issue even begin to become a problem next 3-4 years.
And as it becomes a problem, companies would organically slowly create mentorship and growth programs. The same ones we were running during Covid times where we had not enough engineers and focused on growing junior talent. Companies would react to market conditions start these when there is a need basically.
I’m a head of engineering for a big org. And if I tried to start a program today, I wouldn’t be able to answer “what business need/problem solving with this investment” question I would get from the business leaders as of today. It’s just a long term risk, and we don’t even know if it will pan out or in what ways if it did, what skills to focus on etc... all of these are too early to bet on.
But to answer you question, the skills we will need are probably like; * reading code over writing code * making sure CS fundamentals are still a focus and folks look behind the covers * architecture skills * product understanding, customer understanding, being able prioritize features themselves * software development life cycle skills (TDD, spec driven development etc…) * last 20% software development skills
Am I right? We will see when the day comes.
beefyweefles@reddit
this is precisely what I’d expect an engineering leader to say based on my experience in a number of companies, meanwhile their MO is essentially how much can they hollow out things without it blowing up in their faces
t-tekin@reddit
I think you’re putting a little too much intent and conclusions behind what I said.
There’s definitely a version of this where companies use AI as an excuse to hollow out junior roles. I don’t think that’s healthy, and I don’t think it’s sustainable.
My point was more boring and practical: today, with the market full of strong senior engineers, it is hard to justify a large proactive investment in junior growth purely as a business priority. Not because juniors don’t matter, but because most companies don’t fund long-term pipeline risk until the risk becomes closer, better than understood and more measurable.
That’s not me saying it’s ideal. It’s me saying how these decisions usually get made.
I actually agree with the underlying concern. If we don’t create new paths for juniors to become seniors, we eventually create a serious talent problem. I just think the industry is more likely to react to that problem once it starts hurting hiring, retention, cost, or velocity.
The better version of leadership here is probably not “hire juniors and let them do old routine tasks AI can now do.” It is to redesign junior growth around higher-leverage skills: reading and understanding code, debugging, testing, fundamentals, architecture, product judgment, communication, and learning how to ship production-quality software end to end.
beefyweefles@reddit
I guess from my perspective management seems to not only fail to lead “successful” business outcomes they often fail at really basic technology implementations. So the idea that they can’t do X or Y because of AI has “changed everything” is silly. They’ve always wanted to outsource quality products, quality talent and quality processes with inferior alternatives.
t-tekin@reddit
Fair enough.
We’ve each made our case. I won’t pretend bad management isn’t out there or that big narratives never get used as cover for cost-cutting; they do. And unsuccessful companies don’t exist.
Where I land is just that “we all have to deliver customer value and impact at acceptable quality and cost”. That’s the actual job of engineering.
There are many tradeoffs many orgs indeed do, some poorly by losing customers, but some are extremely successful as well.
And with AI many of that tradeoffs changes drastically as well.
Appreciate the back-and-forth.
beefyweefles@reddit
I'm saying there's an industry standard and groupthink that demands certain things and beats itself up and enforces a certain insanity where they'll do anything but just delivering a genuine product without skimping.
They end up spending way more chasing the ideal of "cost savings" instead of just doing the sensible way and trusting a certain level of expertise. The push for "AI" is a well-established and obvious dynamic of everything-but-substance but on steroids.
BraveResearcher3037@reddit
It won’t help.
The average developer especially junior developer only sticks around for around 3 years. They often doing “negative work” and taking time from seniors.
What is going to happen is when that former junior becomes a mid level capable developer and ask for 25% more money, which is not unreasonable, your HR department is going to push back. They are going to jump ship. Then when it’s time to hire someone new, you are going to have to pay them market wages. If they do stay, they are going to be getting paid less than people you bring in from the outside with the same experience. See “salary compression and inversion”.
This happens across the industry including BigTech companies. Thats why everyone knows you are better off “boomeranging” even if you do like the company you work for.
On the other hand, why wouldn’t I just hire all mid level and senior developers and poach your former junior developer that you have already trained? It’s not like I couldn’t just put an open req on LinkedIn and get hundreds of applications in a day especially if the position were remote.
Yes it’s a collective action problem. But it’s not my problem. My incentives, compensation package and reviews are around me meeting my quarterly and yearly goals not establishing a pipeline for some future manager when I’ll be long gone.
I’m not in a hiring position now. But when I was in a position where I had one or two positions I could fill, I needed someone who I could just throw a problem at that was highly ambiguous and trust them to take ownership of the problem. Anyone who couldn’t do that was a net negative to me and the organization.
razza357@reddit
But isn't a senior's job to mentor, guide and teach juniors? If so, then how is doing so 'taking time away from seniors?
BraveResearcher3037@reddit
There is a middle ground. The mid level developer. The mid level developer is expected to be able to do a task where the business requirements are well defined and use their experience to turn those requirements into code.
The senior’s job becomes to disambiguate, coordinate, code review, deal with the “business”, etc and spend some time coding
In this model, who needs junior developers?
razza357@reddit
I understand your point. The juniors originally used to do the easy grunt work no one else wanted to do, and eventually upskilled to do harder things. That pipeline is broken.
superdurszlak@reddit
Routine kills growth, and did back in 2016-17 when I entered the job market. Gaining different perspectives, running into diverse problems and seeing with your own eyes what works and in what context is what matters.
Doing the same thing over 10 years doesn't mean you have 10YoE, it means you have 1YoE times 10.
Focus on problem solving, rather than becoming a master of holding a particular hammer with a red handle and mastering your grip for 10 years.
Cartindale_Cargo@reddit
Same with path to staff. I was literally working with my manager to figure out what I needed to do for a promotion to staff and now no one knows
BraveResearcher3037@reddit
Senior to staff has nothing to do with AI. Staff is more concerned with organizational strategy. Even Senior can’t be replaced with AI as long as your company doesn’t define Senior as “codez real gud while pulling well defined tickets off a Jira board@
Cartindale_Cargo@reddit
Hah.it does now. Im literally looking at job definitions change. Don't bullshit and say AI has nothing to do with it. The whole industry is changing
BraveResearcher3037@reddit
What do you think the responsibilities of a staff engineer (which I currently am) does that can be duplicated by AI?
Cartindale_Cargo@reddit
I didn't say they could be. I said upper management and c suite are changing the criteria because they all are obsessed with AI. Whether or not it actually matters.
BraveResearcher3037@reddit
And you still didn’t answer the question, “what is the criteria that they are changing the requirements to for a staff engineer”?
Cartindale_Cargo@reddit
I told you. Unknown. The career ladders are unmade and waiting to be built as c suite sits on their ass figuring out what they want their new standards are.
This my main complaint. I have no avenue of actual path atm because the previous set lath no longer exists because everyone wants ai part of it
BraveResearcher3037@reddit
This is your localized company’s problem and you are acting as if this is an industry problem. I can guarantee you that any well known tech company - and even my not well known problem of 1000 employees have a documented leveling guideline for staff.
drguid@reddit
I don't think AI was the reason. Back in 2023 companies just seemed to stop hiring juniors. Why? For a bit more money they could hire a mid-level with tonnes of experience. Also companies are generally only hiring candidates that already can do X, Y and Z. Training is dead and had been for years.
mllv1@reddit
It’s simple: only senior engineers are allowed to use AI
t-tekin@reddit
I don’t understand your take, maybe explain more.
If AI is an important skill to learn,
By limiting the usage to seniors, you literally kill the juniors growth and learn that skill.
IMO what you are proposing is not a solution for junior growth,
this is just protecting your company from the wrong usage of AI because of lack of guardrails.
mllv1@reddit
Using AI is not a skill, that’s the entire point. It’s a big red easy button on your desk. The only reason senior engineers exist to begin with is because that button didn’t exist until recently. Now that it exists, how are juniors supposed to become seniors?
t-tekin@reddit
Here is a simple observed reality that showcases why what you are saying is not correct.
We are seeing how much LLMs amplify the output is extremely different from person to person. (For some it’s 10x impact difference and for some it’s actually degradation) And when we look behind the covers the only difference is how they are used.
So it’s absolutely not a big red button.
Teh_Original@reddit
Impact is not a good metric to sell AI. A meteor has impact. Is a larger impact better?
Illustriousfart@reddit
Sure, but that just widens the productivity gap between junior and senior. Inevitably making it less desirable and economical to hire early career engineers. That's what I want to avoid because it doesn't feel sustainable.
FlowOfAir@reddit
As a mid-level engineer, I've been lately trying really hard to learn and close all gaps to get to the senior level. I think I've learned a few things that could've been vital early on.
That, off the top of my mind. And none of this needs AI.
zxyzyxz@reddit
How do you know or ensure the juniors won't just use AI to write those design docs in the first place?
FlowOfAir@reddit
My take is, it doesn't matter. I too use AI to write design docs, but AI is just really bad at giving you a complete, decent pass on the first attempt. If they don't review the doc themselves, others will find those glaring errors. And a manager is 100% going to tell them "review your stuff first before turning it in" if it's too obvious. Any moderately decent junior engineer should be able to review the doc so it states their intent accurately.
BraveResearcher3037@reddit
I am far away from a junior and I use AI to write design docs. I create my own architecture diagrams well labeled by hand, dump it into ChatGPT, ask it to do a design document and I just keep telling it where it’s wrong, give it more context and keep iterating.
Yes I know how to “work backwards” (how do you say where you worked without saying where you worked) from the business case -> architecture -> implementation -> epics -> workstreams -> tasks.
I don’t care if someone uses AI to write design docs as long as it is coherent, well edited, and they can explain the tradeoffs and their thought process. A lot of what I put into AI is a rambling brain dump.
JuiceAccomplished241@reddit
Yup, AI has shifted the focus from the “what” to the “why” .
Design twice, implement (hopefully) once and they should be able to back every line of code with an explanation.
positev@reddit
And every explanation for a line of code should be recorded in a test communicating the necessity of that kind
Fidodo@reddit
Top of my list of skills to teach would be software design and architecture and debugging. For hiring I always hire for curiosity and that trait has stayed the best signal for a very long time.
For people trying to progress themselves I would say that honestly, nobody really knows what the new software development workflows should be yet. There's a lot of experimentation and a lot of it disagrees with each other and won't pan out or be long lived. Don't be a follower and don't buy into any one approach. If you can be a leader and lead explorations you should do well for yourself.
LetterBoxSnatch@reddit
Yes! Extremely well said. Genuine curiosity; the desire to know mixed with the desire to do, each in balance with the other.
disposepriority@reddit
There is no need for change, or else current juniors would already be seniors with AI, no?
Since that is not the case, the current juniors should follow the same path if they want the same results.
Illustriousfart@reddit
The challenge is that developing a junior into a senior was typically cultivated through progressively more difficult problems. Along this progression people develop the communication and problem solving skills that make experienced engineers what they are.
With AI:
These simple tasks can be orchestrated far faster by senior engineers. The economics of having a junior engineer doing these doesn't seem to work out as they're often not able to work as independently.
Or
We purposefully isolate simpler tasks and assign them to junior engineers. They will likely have a tendency to rely and AI and never learn the fundamentals.
notmsndotcom@reddit
They’ll learn the fundamentals…it’s just the fundamentals today are arguably different. We’ve elevated a level of abstraction so the fundamentals will be more around writing TDDs/specs, soft skills, etc and less about go fix this one line bug by hand
mechkbfan@reddit
Fundamentally they'll never progress past Senior if they just use AI
Same with if you just kept copying and pasting code from Stack Overflow
Or posting for help on Experts Exchange without thinking
Or maybe they'll have the appearance if senior, get promoted and move on, then when someone comes to pick up the pieces after they've left, the business will be asking "why does production break every time a change is made?"
AfricanTurtles@reddit
Your 2nd point is what I have observed. Project managers in their haste to finish things faster and look good at their job will assign things to junior devs and let them run wild with AI. Junior doesn't learn anything, but PM still thinks they're a genius for spitting things out fast.
EyesOfAzula@reddit
I really think Junior developers should focus on being able to speak to the code, regardless of who made it whether it's them or AI. Just saying AI said so is not a good look.
Also, they should have a skeptical mindset. Just like you code review a human dev's work, they should take some steps to code review or validate an agent's work to know when the agent is wrong or off.
I'd say that that's the difference between a vibe code and a software engineer in the AI era.
Livid_Conversation59@reddit
I think about how we hire and develop junior talent so we don't hollow out our senior engineer pipeline. I've been mentoring junior devs recently and noticed that when they're given AI powered tools, they often focus on configuring settings rather than understanding the underlying architecture or design principles. What if teams restructured their work to include more hands on, guided exploration of AI driven solutions? For example, have junior devs help refactor a project using an AI assisted code analyzer.
secondgamedev@reddit
I just ask them to confirm if they can come up with a better solution to the AI answer. It’s forces them to understand the code and see if they can even explain why it’s good. If they think AI solution is best and they can explain it then they can learn from the solution. But mentoring is still kind of required, people don’t know what they don’t know.
bystanderInnen@reddit
Its not distrubting it, its fundmentally changing the Job shifting the skills needed from memorizing synthax to orchestrating/creative Problem Solving.