So far, I'm really struggling to see the value add of a junior engineer given how good the models have become. A senior with Max plan beats three junior hires (very rough estimate).
Having spoken with a few peers (founders), they have expressed similar feelings. I want to hire juniors and see them grow, but I feel my bar of expectations is at an all time high, and perhaps unrealistic.
If you have seniors doing the job of juniors, you will soon have no seniors either. If you have a senior doing the job of 3 juniors (which is different than doing 3x kind of tasks), then your senior will burn out and leave as well.
Also, as senior I enjoy working with juniors because I like to teach and share (with seniors that’s also possible but happens way less often). Working in a team of full senior+ engineers is not fun (in regular companies I mean): too much ego, politics, bias… no open to change
Of course it is true that some juniors outperform competent seniors. It’s also true some seniors are just checked out. But when you are willing to hire folks with meaningful years of experience, you get meaningfully better velocity and quality. It doesn’t matter how simple or complex the problem is.
Those are generalizations but in my experience, once the 5-Year mark passes, a junior is pretty much always a better colleague than a senior.
This is not to say that you should not hire seniors but you need a good mix of trainers and trainees. Training employees can also make a senior much more effective at his own job.
What juniors are sticking around at 5 years? The market reality is they can likely 1.5x-2x their pay by switching jobs, unless they started at FAANG or some other ultra high paying co
And a good junior will ask questions in a design/architecture session that an AI likely wouldn't. This is a good thing to get the group thinking. It also seems that people tend to have better business context than the AIs.
If you’re the type of person or company that would hire an outsourced team overseas without hesitation, then LLMs would probably be fine for you.
The future of OpenAI and Anthropic is contingent on a lot of factors which can (and maybe will) just vanish overnight. If China "invades" Taiwan, or if the data center buildouts completely stall, AI is going to be in a very precarious spot.
A single staff-level engineer with 4 concurrent agents on a Max plan replaces 4 seniors and 12 junior hires.
It's okay to say that we want don't need to hire any human programmers anymore since they are the bottleneck. But the strong disadvantage is that these agents need humans to tell them what to do.
If you think otherwise, look no further than the Huggingface hack and what happens when they run completely autonomously. Not only that is a waste of tokens, it can cause you to be legally liable for damages if something terrible happens.
First, to me, the two major differences between a junior and a senior developer are(or used to be, I'll get to that in a bit):
1. Intuition: software is all about compromise and knowing when you can, should and should not cut corners and at what cost really is, to a very high degree, intuition(and some auxilary technical knowledge). 2. Domain knowledge.
Both come from, you guessed it, experience.
The second part is junior developers in general - from what I see, they are commonly completely detached from reading and writing code. Currently Gen-Z are entering the workforce. For many of them(vast majority), cut off their connection to the internet and they struggle with a simple while loop. And they look horrified when I tell them that 15 years ago, we were writing code - SVMs on a sheet of paper as exams. Same goes for all the people who said "oh well, no need to write the code if the AI can do it for me". Seeing people with two decades of experience turn into idiocracy characters is painful. Like in a year I see the seniors revert back into what juniors are today - completely incapable of formulating and facing a problem without an LLM on their side. And worse still - they will never admit it but I see how much they stop being passionate and just roll with everything.
The other thing to consider is burnout: take a senior that understands your service or product completely: in a real world scenario you know that no system is ever going to be perfect. You can only push someone so far with mountains of small, medium and large tasks and tell them "what's the big deal, you have unlimited slop max pro"? The big deal is small and trivial tasks which, with or without AI, require context switching which is exhausting. It's not a question of "if" but "when" will your senior get burned out and toss everything at slop max pro++ and just collect a paycheck for as long as the company is around or something better pops up. And in that same context - if you come to me with 5 jira tickets and tell me they have to be done by the end of the day, do I really have the time and mental strength to pay close attention to all of them or just be like "hey slop, see this ticket? do it? when tests pass, deploy". At this point, my question is, am I a senior developer if I was practically never in the loop of what was done?
And yeah... have seen all the jargons come and go. DOS, Novell, Windows NT, xml, soap, next, BSD had a moment, chef and yada yada...the vague hand wavy semantics to 2010s web saas will not be spared
The most common shared cross continent language for whats going on is math... stats of state of memory registers
The jargons conjured up on the fly to do office work 30+ years ago all fade away; how many startups went all in on COBOL during the RoR era?
So yeah; hard pressed to recommend tethering juniors to useless jargon as we have most of my career.
I think it all comes down to what the they did during their studies and how they've spent their time. So I'd judge them based on the things they've built and their experience using AI tools. These days you literally have top US companies like Sierra asking their candidates to vibe coding a project as the coding interview.
I personally spent a lot of time and money building with Claude Code and Cursor. Probably more than you would at most companies, especially an intern: full days (weekends included) and about $5,000 in 2 years on tokens. During my sophomore year, I went so hard building agents with AI, I skipped all my classes and barely passed my subject, which got me Tier-1 VC inbounds. So the junior engineers in their early 20s can be more proficient than senior devs.