AWS CEO says replacing junior devs with AI is 'one of the dumbest ideas'

I strongly disagree, a Senior who cannot ask a "dumb question" is a useless developer to me. Ask away, we want to figure things out, we're not mind readers. Every good senior developer I know asks questions and admits when they don't know things. This is humility and in my eyes is a strong market for a good Senior Developer. The entire point of our job is to ask questions (to ourselves typically the most) and figure out the answers (the code).

Juniors could or should ask more questions, and by the time you're a Senior, you're asking key questions no matter how dumb they sound.

This seems almost entirely wrong to me? That anyone, at any level of seniority, can ask "dumb questions" and give signal about "nonsense abstractions" seems a property of any healthy organization. That only juniors can do this doesn't just seem wrong, it seems backwards. I would expect seniors to have the clearest idea on whether abstractions make sense, not juniors.

Junior devs do that naturally (if you have the culture) because they don’t know anything already. It’s great

One is a "one junior per team" model. I endorse this for exactly the reasons you speak.

Another, as I recently saw, was a 70/30 model of juniors to seniors. You make your seniors as task delegators and put all implementation on the junior developers. This puts an "up or out" pressure and gives very little mentorship opportunities. if 70% of your engineers are under 4 years of experience, it can be a rough go.

Juniors are usually given either grunt or low priority work while seniors get more "important" work.

OTOH, it takes a lot to get your questions on RIGHT EARS when you're a junior, so wouldn't agree with your characterization at all.

Really, juniors are only important because they ask "dumb" questions that can help remove useless abstractions? That your take?

It’s damned near impossible to figure out where to spend your time wisely to correct an assumption a human made vs. an AI on a blended pull request. All of the learning that happens during PR review is at risk in this way and I’m not sure where we will get it back yet. (Outside of an AI telling you - which, to be fair, there are some good review bots out there)

Juniors are just... necessary in the balance, have to little of them and the mid and senior devs will get more and more expensive, so you hire a bunch of juniors, they get trained on job, and it balances it out.

Hell, if company does it right they might underpay junior-turned-senior for decade before they notice and look how industry pay looks like!

I still think the central issue is the economy. There are more seniors available to fill roles, so filling out junior roles is less desirable. And perhaps "replacing juniors with AI" is just the industry's way of clumsily saving face.

Or in other words, the people that get the most value from AI are junior devs, since they still don't know very well plenty of popular environments. It's also useful for seniors that are starting something in a new environment, but those only get 1 or 2 novel contexts at a time, while everything is new for a junior.

Or, in again another set of words, AI enable juniors to add more value more quickly. That makes them more valuable, not less.

I also don’t believe juniors, kids, seniors, staff, principals, distinguished/fellow should be replaced by AI. I think they WILL be, but they shouldn’t be. AI at Gemini 3 Flash / Claude Opus 4.5 level is capable with help and review of doing a lot of what a lot of devs do currently. It can’t do everything and will fail, but if the business doesn’t care, they’ll cut jobs.

Don’t waste time trying to argue against AI to attempt to save your job. Just learn AI and do your job until you’re no longer needed to even manage it. Or, if you don’t want to, do something else.

> cargo-culting Stack Overflow

What do you mean by this? I understand “cargo-culting” as building false idols, e.g. wooden headphones and runways to attract airplanes that never come.

This is the only role of executives, sales people, account managers. They usually do it with complete and utter confidence too. Vibe-questioning and vibe-instructing other people without a care in the world.

AI is now just the scapegoat for an economy wide problem. Execs found "one neat trick", piling junior work on seniors until they quit. While not hiring replacements in order to goose short term profits. Now every company is in the same position where hiring a senior really means hiring 5 seniors to replace the one that had 5 jobs layered on over a few years. This is of course impossible for any mortal to jump into. Now they also dont even have juniors to train up to senior levels.

I was never formally trained so I just keep asking "why" until someone proves it all the way. Sales itself is also a lot about asking questions that won't come up to find the heart of the thing people actually want... which is just another side of the coin.

Now that their minds are free from routine and boilerplate work, they will start asking more 'whys' which will be very good for the organization overall.

Take any product - nearly 50% of the features are unused and it's a genuine engineering waste to maintain those features. A junior dev spending 3 months on the code base with Claude code will figure out these hidden unwanted features, cull them or ask them to be culled.

It'll take a while to navigate the hierarchy but they'll figure it out. The old guard will have no option but to move up or move out.

I would argue a machine that short circuits the process of getting stuck in obtuse documentation is actually harmful long term...

But for many Jr engineers it’s the hard part. They are not (yet) expected to be responsible for the larger issues.

How are we defining "learning" here? The example I like to use is that a student who "learns" what a square root is, can calculate the square root of a number on a simple 4 function calculator (x, ÷, +, -) if iteratively. Whereas the student who "learns" that the √ key gives them the square root, is "stuck" when presented with a 4 function calculator. So did they 'learn' faster when the "genie" surfaced a key that gave them the answer? Or did they just become more dependent on the "genie" to do the work required of them?

This really isn't the case from what I've seen. It's that they use Cursor or other code generation tools integrated into their development environment to generate code, and if it's functional and looks from a fuzzy distance like 'good' code (in the 'code in the small' sense), they send an oversized PR, and it's up to the reviewer to actually do the thinking.

Ehh... 'used well' is doing some very heavy lifting there. And the incentive structure at 90% of companies does not optimize for 'using it well.'

The incentive is to ship quickly, meaning aim the AI-gun at the codebase for a few hours and grind out a "technically working" solution, with zero large-scale architecture thought and zero built-up knowledge of how the different parts of the application are intended to work together (because there was no "intention"). There will be tests, but they may not be sensible and may be very brittle.

Anyway, deploying a bunch of fresh grads armed not with good mentorship but with the ability to generate thousands of LOC a day is a recipe for accelerating the collapse I usually see in startup codebases about 6-8 years old. This is the point where the list of exceptions to every supposed pattern is longer than the list of things that follow the patterns, and where each bug, when properly pursued, leads to a long chain of past bad decisions, each of which would take days of effort to properly unwind (and that unwinding will also have a branching effect on other things). Also, coincidentally, this is the point where an AI agent is the most useless, because they really don't expect all the bizarre quirks in the codebase.

Am I saying AI is useless? No, it's great for prototyping and getting to PMF, and great in the hands of someone who can read its output with a critical eye, but I wouldn't combine it with inexperienced users who haven't had the opportunity to learn from all the many mistakes I've made over the years.

I hate to be so negative, but one of the biggest problems junior engineers face is that they don't know how to make sense of or prioritize the gluttony of new-to-them information to make decisions. It's not helpful to have an AI reduce the search space because they still can't narrow down the last step effectively (or possibly independently).

There are junior engineers who seem to inherently have this skill. They might still be poor in finding all necessary information, but when they do, they can make the final, critical decision. Now, with AI, they've largely eliminated the search problem so they can focus more on the decision making.

The problem is it's extremely hard to identify who is what type. It's also something that senior level devs have generally figured out.

This is "the kids will use the AI to learn and understand" level of cope

no, the kids will copy and paste the solution then go back to their preferred dopamine dispenser

Since desktop computers became popular, there have been thousands of small to mid-size companies that could benefit from software systems.. A thousand thousand "consultants" marched off to their nearest accountant, retailer, small manufacturer or attorney office, to show off the new desktop software and claim ability to make new, custom solutions.

We know now, this did not work out for a lot of small to mid-size business and/or consultants. Few could build a custom database application that is "good enough" .. not for lack of trying.. but pace of platforms, competitive features, stupid attention getting features.. all of that, outpaced small consultants .. the result is giant consolidation of basic Office software, not thousands of small systems custom built for small companies.

What now, in 2025? "junior" devs do what? design and build? no. Cookie-cutter procedures at AWS lock-in services far, far outpace small and interesting designs of software.. Automation of AWS actions is going to be very much in demand.. is that a "junior dev" ? or what?

This is a niche insight and not claiming to be the whole story.. but.. ps- insert your own story with "phones" instead of desktop software for another angle

"AI" tools are most useful in the hands of experienced developers, not juniors. It's seniors who have the knowledge and capability to review the generated output, and decide whether the code will cause more issues when it's merged, or if it's usable if they tweak and adapt it in certain ways.

A junior developer has no such skills. Their only approach will be to run the code, test whether it fulfills the requirements, and, if they're thorough, try to understand and test it to the best of their abilities. Chances are that because they're pressured to deliver as quickly as possible to impress their colleagues and managers, they'll just accept whatever working solution the tool produces the first time.

This makes "AI" in the hands of junior developers risky and counterproductive. Companies that allow this type of development will quickly grind to a halt under the weight of technical debt, and a minefield of bugs they won't know how to maneuver around.

The unfortunate reality is that with "AI" there is no pathway for junior developers to become senior. Most people will gravitate towards using these tools as a crutch for quickly generating software, and not as a learning tool to improve their own skills. This should concern everyone vested in the future of this industry.

Because that makes the most business sense.

https://substack.com/@kentbeck

What software projects is he actively working on?

Sorry, what does that mean exactly ? Are you claiming that a junior dev knows how to ask the right prompts better than a Senior dev ?

For medium or small companies, these guardrails or documentation can be missing. In that case you need experienced people to help out.

Me too. Fire your senior devs. (Ha ha, not ha ha.)

While anyone is free to define words as they so please, most people consider those with the most experience to be seniors. I am pretty sure that has been the message around this all along: Do not cut the seniors. The label you choose isn't significant. Whether you want to call them juniors or seniors, it has always been considered to make no sense to cut those with the most experience.

I think LLM is a reflection of human intelligence. If we humans become dumber as a result of LLM, LLMs will also become dumber. I’d like to think in some dystopian world, LLM’s trained from pre 2023 data will be sought after.

I would have agreed with you 100% one year ago. Basically senior engineers are too complacent to look at AI tools as well as ego driven about it, all while corporate policy disincentivizes them from using anything at all, with maybe a forced Co-Pilot subscription. While junior engineers will take a risk that the corporate monitoring of cloud AI tools isn't that robust.

But now, although many of those organizations are still the same - with more contrived Co-Pilot subscriptions - I think senior engineers are skirting corporate policy too and become more familiar with tools.

I'm also currently in an organization that is a total free for all with as many AI coding and usage tools as necessary to deliver faster. So I could be out of touch already.

Perhaps more complacent firms are the same as they were a year ago.

Coding in any sufficiently large organization is never the main part of senior's time spend, unless its some code sweatshop. Juniors can do little to no of all that remaining glue that makes projects go from a quick brainstorming meeting to live well functioning and supported product.

So as for worth - companies can, in non-idedal fashion obviously, work without juniors. I can't imagine them working without seniors, unless its some sort of quick churn of CRUDs or eshops from some templates.

Also there is this little topic that resonates recently across various research - knowledge gained fast via llms is a shallow one, doesn't last that long and doesn't go deeper. One example out of many - any time I had to do some more sophisticated regex-based processing I dived deep into specs, implementation etc. and few times pushed it to the limits (or realized task is beyond what regex can do), instead of just given the result, copypasted it and moved along since some basic test succeeded. Spread this approach across many other complex topics. That's also a view on long term future of companies.

I get what you say and I agree partially but its a double edged sword.

Nothing is preventing you from studying how the bugfix works once it's in place.

Nor is there any reason this use of AI should cause you to lose skills you already have.

That's my thought too. It's going to be a triple whammy

1. Most developers (Junior and Senior) will be drawn in by the temptation of "let the AI do the work", leading to less experience in the workforce in the long term.

2. Students will be tempted to use AI to do their homework, resulting in new grads who don't know anything. I have observed this happen first hand.

3. AI-generated (slop) code will start to pollute Github and other sources used for future LLM training, resulting in a quality collapse.

I'm hoping that we can avoid the collapse somehow, but I don't see a way to stop it.

The thing with juniors is: those who are interested in how stuff works now have tools to help them learn in ways we never did.

And then it's the same as before: some hires will care and improve, others won't. I'm sure that many juniors will be happy to just churn out slop, but the stars will be motivated on their own to build deeper understanding.

It should probably be supplemented with some good old RTFM, but it does get us somewhat beyond the "blind leading the blind" StackOverflow paradigm of most software engineering.

Even more recently we had this with radiologists, a profession that was supposed to be crushed by deep learning and neural networks. A quick Google search says an average radiologist in the US currently makes between $340,000 to $500,000 per year.

This might be the professional/career version of "buy when there's blood in the streets."

Hell, I should probably be studying how to be a carpenter given the level at which companies are pushing vibe coding on their engineers.

That's such a terrible trend.

Reminds me of my peers back in ~2001 who opted not to take a computer science degree even though they loved programming because they thought all the software engineering jobs would be outsourced to countries like India and there wouldn't be any career opportunities for them. A very expensive mistake!

Historically, these candidates have been the hiring sweet spot: less risky than brand new engineers, still small enough experience to efficiently mold them into your bespoke tools and processes and turn them into long-term employees, and still very cheap.

It's similar to all those people who were hyping up blockchain/crypto/NFTs/web3 as the future, and now that it all came to pass they adapted to the next grift (currently it's AI). He is now toning down his messaging in preparation of a cooldown of the AI hype to appear rational and relevant to whatever comes next.

https://www.business-standard.com/amp/world-news/amazon-euro...

Considering the talk around junior devs lately on HN, there's way too many of them, it would indeed be amusing.

To what?

You can describe pre-ai developers and like that too. It's probably my biggest complaint about some of my Co workers

In my view there's two parts to learning, creation and taste, and both need to be balanced to make progress. Creation is, in essence, the process of forming pathways that enable you to do things, developing taste is the process of pruning and refining pathways to doing things better.

You can't become a chef without cooking, and you can't become a great one without cultivating a taste (pun intended) for what works and what it means for something to be good.

From interactions with our interns and new-grads, they lack the taste, and rely too much on the AI for generation. The consequence is that when you have conversations with them, they straggle to understand the concepts and tools they are using because they lack the familiarity that comes with creation, and they lack the skills to refine the produced code into something good.

It's not really the work that LLMs currently do. I mean sure, maybe if you plug an LLM to read all emails and slacks and zoom transcripts of the entire company, it could do it at some point in the future. But would it have the same amount of influence compared to an industry & company veteran who has the company specific knowledge and experience that is nowhere written down?

Why? That seems unlikely to me. That's like saying juniors are likely the most comfortable with jj, zed, or vscode.

Week 2: 32 interns

Week 3: 16 interns

Week 4: 8 interns

Week 5: 4 interns

Week 6: 2 interns

Week 7: 1 intern

Week 8: 0.5 interns

Is it possible to make it to the end of the summer without getting sliced in half?

What the hell.

Consider making them fight each other in an arena, you could monetize that.

I see none of that happening - software quality is actually in freefall (but AI is not to blame here, this began even before the LLM era), delivery doesn't seem to be any faster (not a surprise - writing code has basically never been the bottleneck and the push to shove AI everywhere probably slows down delivery across the board) nor cheaper (all the money spent on misguided AI initiatives actually costs more).

It is a super easy bet to take with money - software development is still a big industry and if you legitimately believe AI will do 90% of a senior engineer you can start a consultancy, undercut everyone else and pocket the difference. I haven’t heard of any long-term success stories with this approach so far.

I'm yet to see that production-grade code written by these production-grade models;