My take on how a decade (or more) of using cloud services for everything has seemingly deskilled the workforce.
Just recently I found myself interviewing senior security engineers just to realize that in many cases they had absolutely no idea about how the stuff they supposedly worked with, actually worked.
This all made me wonder, is it possible that over-reliance on cloud services for everything has massively deskilled the engineering workforce? And if it is so, who is going to be the European clouds, so necessary for EU’s digital sovereignty?
I did not copy-paste the post in here because of the different writing style, but I get no benefit whatsoever from website visits.
I think its actually that most people generally don’t really understand most things beyond the minimal level necessary to get by. Now that the tech industry isn’t just a bunch of nerds you’re increasingly more likely to encounter people who are temperamentally disinclined to seek understanding of those details.
I went through hiring several times at several companies, being on the interviewer side.
Typically it’s not the talent pool as much as what the company has to offer and how much they’re willing to pay. I referred top notch engineer friends, and they never made it past HR. A couple were rejected without interview because they asked too high of a salary, despite asking under market average. The rest didn’t pass HR on personnality or not having all the “requirements”, because the really good engineers are socially awkward and demand flexibility and are honest on the résumé/CV, or are self taught and barely have high-school graduation on there (just like me).
I’ve literally seen the case of: they want to hire another me, but ended up in a situation where: I wouldn’t apply for the position myself, and even if I did, I wouldn’t make it to the interview stage where I’d talk to myself and hire myself.
Naturally the candidates that did make it to me weren’t great. Those are the people that do the bare minimum, have studied every test question (without understanding), vibe code everything, typically on the younger and very junior side. They’re very good at passing HR, and very bad at their actual job.
It’s not the technology, it’s the companies that hire that ultimately steers the market and what people study for. Job requirements are ridiculous, HR hires engineers on personnality like they’re shopping for yet another sales associate, now it takes 6 rounds of interviews for an entry level position at a startup. VC startups continue to pay wildly inflated wages to snatch all the top talent while established companies are laying off as much IT staff as possible to maximize profits.
I have the opposite experience, when I was doing interviews I just skipped the very obviously underskilled people (which, IIRC were in the single digits) and interviewed pretty much everyone.
For context, I’m the main architect and dev of the company I was hiring for. Most of the candidates were horrible.
A lot of this has to do with recruiters. I’ve been interviewing for a few years at my company such with as many different sets of recruiters, from recruiting firms to our corporate recruiters, to ones we hired ourselves. Our corporate recruiters handed over garbage candidates who we could often tell wouldn’t work out after the first 10 min of the interview, whereas the other two groups of recruiters would do a good job filtering so we’d get than a 50% hit rate on our first round. Unfortunately, we promoted our recruiters once the need for talent dropped (or they moved on to a recruiter firm), and now they’re unwilling to go back to recruiting.
The quality of your recruiter matters quite a bit, so you’ll want to find someone who is experienced hiring a certain type of person so they know what to look for.
I totally agree with you, but I don’t think this is the specific case. Most of the rejections in our case (which I can see) on the preliminary screening were based on lacking CV skills. Which is stupid in its own way, but at least makes sense assuming we are looking for those skills specifically.
For the rest, the company is a remote company paying good salaries for the European market, I would say slightly above market average in many metrics.
I will sift more into the rejections, but from what I have seen, almost all those who had the screening phone call made it to the interview (I.e., rejections were mostly cv-based).
The main factor, IMO, is that everyone wants good engineers but good engineers don’t change jobs that often.
Meaning most of the candidates you interview will suck in one way or another.
And everyone calls themselves “senior” nowadays.
Exactly. We don’t hire “junior” positions, because all the midlevels are juniors, all seniors are mid-level, and seniors don’t apply. I’m a senior and a recruiter found me, I didn’t apply (at least not to this company).
Everyone calls themselves senior because that’s the only type of position recruiters look for.
I’m a mid level dev, but I’m encouraged by recruiters to apply for senior positions because their clients are actually looking for a range of levels
Yeah, that’s true, everyone thinks they want a senior where usually someone who’s not a straight up junior is more than enough. And a fast learning and motivated junior is the best you can get, IMO, though those are pretty rare as well.
I’m reminded of when my boss asked me whether our entry test was too hard after getting several submissions that wouldn’t even run.
Sometimes prospective employees are just shit.
Ours is really simple, like something any somewhat competent engineer could complete in half the time we provide after going through the tutorial on the framework’s website. Yet so many people fail, even when they claim to have years of experience with the framework.
There are a ton of terrible applicants out there.
I got asked the same. I simply pointed out the test is a reproduction of last week’s bug that took down prod at 2am and got paged to fix, and is therefore as realistic as it gets of what they’ll need to be able to handle.
It’s always DNS, everyone should know that.
It’s always DNS, everyone should know that.
It’s not DNS. There’s no way it is DNS. It’s not technically possible for it to be DNS.
And it’s always DNS.
Ahaha yes, that might be the case, but I started to lose hope if the top of the applicants (out of hundreds of rejected!) all exhibits this behavior. I can’t help but feel that now we are looking for people with a mindset and skillset that is simply disappearing in the industry.
And as I said in another post, I perfectly acknowledge that if I stopped reading and investigating stuff on my own, I could absolutely keep my job by just mindlessly administering a few services and rephrasing CIS benchmarks…
I get what you’re saying, but also see the other side - these services exist and aren’t ever going away, so the level of knowledge you need about these to use them at least competently is significantly reduced.
What their existence does mean is that there are thousands of developers who wouldn’t ever touch or learn any of this stuff previously are now actually learning it and using it. That’s a positive thing. Not everyone needs to be an expert on the inner workings of everything that a service provides unless you’re specifically looking for an expert.
Also……people lie on CVs and cover letters. If your ad has buzzwords and technology X, Y, and Z, then you should expect people with little to no knowledge of at least one of those things to have all 3 on their resume.
I partially agree, but not only we are looking for experts of that thing, we are also looking for security experts, and security knowledge is very much meta-knowledge. A software developer might not care at all about - say - how the CI/CD works, because all they care is that the thing builds the code. A security expert generally has a broader scope, and their job is not functional, which means their job is exactly understanding the thing to be able to model the risks around it. So they might not care of all the tools used in that CI/CD or the exact details of the steps, but they should understand the execution flow, the way third party dependencies are pulled, verified, consumed, the authorization model etc.
There is no such thing of security professional who doesn’t understand - at least from an academic point of view - the overall setup of a thing they worked with.
If I take the image attestation example I made in the post, I consider the “inner workings” to be the cryptographic details, such as ciphers and their working mechanisms, or the exact details of the way that attestation can be verified offline, or what exactly is computed and how. I am OK with someone not knowing this. But not understanding the whole flow? Well, without this what’s left? Copying the 3 lines of code that do something from the Github documentation? Any software engineer can very much do that, what is your contribution as a security specialist?
……people lie on CVs and cover letters. If your ad has buzzwords and technology X, Y, and Z
Totally agree. It is very likely, although the more people I interview, the more I think that they are not lying from their perspective. It’s that people can legitimately make a career today by stitching together stuff with scotch tape, spending years by just by doing that and effectively have little to show for those years. But from their perspective, they might be experienced in that stuff, maybe?
That is technically correct in a way, but I’ll argue very wrong in a meaningful way.
Cloud services are meant to let you focus less on the plumbing, so naturally many skills in that will not be developed, and skills adjacent to it will be less developed.
Buttttt you must assume effort remains constant!
So you get to focus more on other things now. E.g. functional programming, product thinking, rapid prototyping, API stuff, breadth of languages, etc. I bet the seniors you are missing X and Y in have bigger Zs and also some Qs that you may not be used to consider, or have the experience to spot and evaluate.
Mind you that my take and experience is specifically in the context of security.
I struggle to make the parallel that you suggest (which might work for some areas) with a security engineer.
Say, a person learned to brainlessly parrot that pods need to have setting x or z. If they don’t understand them, they can’t offer meaningful insight in cases where that’s not possibile (which might be specific), they can’t provide a solid risk analysis etc.
What is the counterpart to this gap? Because I struggle to see it. Breadth of areas where this superficial knowledge is available is useless, IMHO.
Because a security engineer focused on cloud would rightfully say “pod security is not my issue, I’m focused on protecting the rest of our world from each pod itself.”. With AWS as example: If they then analyze the IAM role structures and to deep into where the pod runs (e.g. shared ec2 vs eks) etc. then it would just be a matter of different focus.
Cloud security is focused on the infrastructure - looks like you’re looking for a security engineer focused on the dev side.
If they bring neither to the table then I’m with you - but I don’t see how “the cloud” is at fault here… especially for security the world as full of “following the script” people long before cloud was a thing.
I mean, the person in question had “hardening EKS” on their CV. EKS still means that the whole data plane is your responsibility. How can you harden a cluster without understanding the foundation of container security (isolation primitives, capabilities, etc.)? Workload security is very much part of the job.
I mean the moment some pod will need to run with some privilege (say, a log forwarder which gets host logs), and you need to “harden” the cluster, what do you do if you don’t understand the concept of capabilities? I will tell you what, because I asked this very question, and the answer was “copy the logs elsewhere”, which is the “make it work with the hammer solution” that again shows the damage of not understanding.
I am with you about different scopes, skillsets etc. But here we were interviewing people with a completely matching skillset on paper.
Oh yeah I see…
As some old philosopher once said: “shit’s fucked, yo”.
Seems to be appropriate here.
Yeah I can see that.
However, you are now arguing a different point than I am getting from your original post. Maybe my fault in interpretation ofc, but the main difference (in my view) is:
You say “incompetent” and “less skilled” as general statements on senior engineers. Those statements are false.
You also say “missing the skills you are looking for” which is obviously true.
And the implication that before cloud, people developed the specific skills you need more naturally - because they had to. This makes sense and I believe it.
You say “incompetent” and “less skilled” as general statements on senior engineers. Those statements are false.
I am saying that the competencies of people who grew up (professionally) with outsourced services are more superficial and give them way less understanding (and agency) on the systems they oversee. I make the opinionated argument that knowing which service to use in a cloud provider is not just a different skill from implementing that functionality “manually”, but is hierarchical inferior, easier to acquire and less useful in general.
A weird parallel would be someone who hikes 100% of the time with a guide who takes care of orientation, camp setting etc., and someone who goes alone. If I am simply comparing the pictures they are showing me, I might not appreciate the difference, but if you asked me who I would trust to come hiking with me, I wouldn’t have doubt, because I consider the skill “finding, choosing and listening to the guide” to be hierarchial inferior to “orient, set camp etc. by yourself”.
So it’s not just a matter of matching the skills I need, is actually a much broader argument about deskilling engineers.
I understand.
Obviously, “knowing which cloud services to enable” is a lesser skill than knowing how those services work. That is not a parallel or equal skill in any way.
But do you assume people are just going drrrrr brain off when they don’t learn that one skillset you are accustomed to spotting?
Well, for the relatively small sample of Kubernetes experts I interviewed, basically any topic beyond “you use this tool” was a disaster, including Kubernetes knowledge. I am not selective, it’s not like I expect a specific skillset, but what would you think if someone with a decade of platform security doesn’t understand cryptography and supply chain, Linux permissions, Kubernetes foundational concepts, container isolation or networking? At some point the question is legitimate, what are you expert in? The answer I have been able to give myself so far is “stitching together services that do stuff” and “recommend what the documentation/standard recommends”. I consider myself satisfied to have somewhat decent knowledge in some of those areas, I am not expecting someone understanding all of that, but none of them? Maybe from someone who just joined the industry.
That being said, I am genuinely frustrated by how little people know or care about the plumbing these days. :D
I am so fucking tired of seeing someone spin up 3 cloud databases for what could be a 40k in-memory hashtable.
I’m a very good engineer, but so much of my time is consumed fighting with Tekton pipelines and migrating testing frameworks and versions I barely have time to write code. But that’s because I can figure that stuff out when I have to. All the code is written by the people who can’t figure that stuff out.
Why this isn’t two separate jobs I can’t understand. Let me do some stuff I’m good at rather than constantly fighting with things I’m not?
This hits the nail right on the head. The point of cloud services is to take away all the overheads of building and delivering software solutions that have nothing to do with the actual business problem I’m trying to solve.
If I want to get a new product to market, I want to spend most of my time making my core product better, more marketable, more efficient. I don’t want to divert time and resources to just keep the lights on, like having to hire a whole bunch of people whose only jobs is to provision and manage servers and IT infrastructure (or nurse a Kubernetes cluster for that matter). Managing Kubernetes or physical tin servers is not what my business is about. All this tech infrastructure is a means to an end, not the end itself.
That’s why cloud services is such a cost efficient proposition for 98% businesses. Hell, if I could run everything using a serverless model (not always possible or cost effective) I’d do it gladly.
This is quite a trite argument from my point of view. Also, this is from the perspective of the business, which I don’t particularly care about, and I tend to look from the perspective of the worker.
Additionally, the cloud allows to scale quickly, but the fact that it allows to delegate everything is a myth. It’s so much a myth that you see companies running fully on cloud with an army on people in platform teams and additionally you get finops teams, entire teams whose job is optimizing the spend of cloud. Sure, when you start out it’s 100% reasonable to use cloud services, but in the medium-long term, it’s an incredibly poor investment, because you still need people to administer the cloud plus, you need to pay a huge premium for the services you buy, which your workforce now can’t manage or build anymore. This means you still pay people to do work which is not your core business, but now they babysit cloud services instead of the actual infra, and you are paying twice.
Cloud exploded during the times of easy money at no interest, where startups had to build some stuff, IPO and then explode without ever turning a single dollar of profit. It’s a model that fits perfect in that context.
At least where I work, our cloud team is ~35 people who manage the whole thing.
The datacenter team? In the hundreds.
Cloud is not the answer to every infra problem, but the flexibility, time to market, and lifecycle burden are easily beneficial weighed against finops. I’m an Azure engineer myself, it’s no comparison the benefits to a managed solution vs rolling your own DC for a lot of regular business workloads and solutions. Beyond that personally I’ve been able to skill up in areas I wouldn’t be able to otherwise if I was stuck troubleshooting bad cables, rebuilding a dead RAID array, or planning VMWare scaling nonsense.
But those are absolutely not the only 2 levels. Server rental can be managed easily by the same infra team who manages the cloud, for a fraction of cost.
I will say more, the same exact team that spends time managing EKS clusters could manage self-managed clusters and have money to spare for additional hires.
I will say more, the same exact team that spends time managing EKS clusters could manage self-managed clusters and have money to spare for additional hires.
Your suggestions is a large expansion of skillset needed for your alternative to the cloud solution. Your own experience in attempting to hire workers should point to the reason thats a bad idea. You’re going to need even higher skilled people, and they are going to ask for significantly more money.
I wouldn’t say it’s a large expansion of skillset, meaning it’s not massive. But yes, indeed it is problematic to find people. It is because this is a vicious circle in which companies are digging their own graves by eliminating a market for those people, which in turn means that those who would want to hire some can’t find them easily, leading to outsourcing instead. Do this for 15 years across the whole industry and it stops being an option, which is pretty much where we are today. That said, training and upskilling is always a possibility for companies who invest on their own employees and are playing the long game…
I get you that it’s easy to over-provision in the cloud, but you can’t return an on-prem server. A cloud VM, just shut it down and you’re done.
AWS talks about minimizing undifferentiated heavy lifting as a reason to adopt managed services and I find that largely to be true. The majority of companies aren’t differentiating their services via some low-level technology advantage that allows them to cost less. It’s a different purchasing model, a smoother workflow, or a unique insight into data. The value an organization provides to customers should be the primary focus of the business, the rest is a means to sharpen that focus.
A cloud VM, just shut it down and you’re done.
If this flexibility is needed, and it’s an “if”, a dedicated server does the same. But even a cloudVM is already lower level compared to other services (which are even more abstract) - like EKS, SQS, etc.
The value an organization provides to customers should be the primary focus of the business, the rest is a means to sharpen that focus.
In my experience this often translates in values that flows to AWS, while the company giving value to customers is stuck with millions of cloud bills each month, and a large engineering footprint that eventually needs to cut, leaving fewer and fewer people working on the product.
That said, I acknowledge that cloud has business reasons to exist, I wrote an entire other post about my hate for it, but I still acknowledge that. However there are some myths that finally are getting dispelled (outsource infra and focus on your product).
I’d like to understand how self managing all the lower level components abstracted by the cloud is saving on headcount. Care to math that out for us?
It depends. An EKS cluster can cost easily 20x what an equivalent cluster costs with same resources. The amount of people necessary to manage it is very close compared to a bare cluster, which depending on the scale can save hundreds of thousands or millions per year, therefore allowing extra headcount.
For example, a company I worked for had a team of 6 managing all their kubernetes cluster on rented dediservers. The infra costed around 50k/year. The same clusters on EKS could be managed by 4 people (maybe?), but would have costed easily 5-600k, especially since they were beefy machines, possibly even more. That amount of money would pay for 7-8 additional headcount in local hires.
Considering that in those clusters there were 40-50 postgres clusters, if moving those to RDS they would have probably looked at millions in cloud bills per year, and the effort to run those dB’s once the manifests were developed was negligible (same team was managing them). This was a tiny startup, with limited resources for internal tools and automation development.
So it’s not like managing everything can save headcount, it’s that not outsourcing everything can save so much money that largely compensates for more headcount, plus you are giving money to real people, who spend local and pay taxes.
I’m not in any way, shape, or form an engineer so I don’t really understand the exact details of your post.
However, you post reminded me of a really good episode of a podcast called Hidden Brain. In it the host, discusses the topic of knowledge with a cognitive scientist.
At one point, they talk about how sophisticated technology has gotten that people don’t know how to solve problems if that technology brakes, especially since technology is getting so good that it makes fewer mistakes. They use an airplane as an example in which an experienced pilot forgot how to get out of a nosedive and crashed the plane. On a smaller scale, the host mentioned that he has a hard time navigating if his phone’s GPS doesn’t work.
Its a really interesting listen if you have the chance.
Thanks, indeed I think there are many parallels with other areas. I will check it out.
Or maybe it’s just a different skill set
Not when the skillset is essentially outsourced and you are left consuming the product of that skillset.
Understanding is nonnegotiable in security, IMHO.
You can’t fail to understand how signature attestation works, if you are implementing it, to make one example I made in the post. Otherwise you end up verifying the signature in the CI (like that person claimed it should be done) and waste the whole effort. You can definitely still outsource the whole infra and scripting to Github, but you still need to understand. The problem is that when you can outsource everything, at some point understanding becomes an extra step.
interviewing senior security engineers
Or maybe senior security engineers from 10 years ago were somewhat different from (wannabee senior) security engineers today?
Did you ask them to write 0xD6 in decimal? 😃
That’s the thing! I think it wouldn’t be conceivable that your “principal engineer” (real position for one of the people) doesn’t understand the basic theory of the stuff they are implementing. Now it feels you can instead work years and years just shuffling configuration and pressing buttons, leading to “senior” people who didn’t gather actual years of experience.
I don’t want to pretend I am outside this logic. I am very much part of this problem myself, having started my career 10 years ago. I do despise cloud services though (if anything, they are super boring), so I tend to work with other stuff. But I could 100% just click buttons and parrot standard and keep accruing empty years of experience…
You want to hire the “guru”, not the “principal”. You want to actually ask him to write 0xD6 in decimal, and if he dares to answer “Seriously? Come on now, that’s boring”, then you hire him on the spot.
But you can’t hire only gurus. You need normal seniors, too. Build a normal team around one guru. Maybe build one ultra advanced team around 2-3 gurus, if you really need to invent new and hardcore difficult stuff.
I agree with your lack of affection for cloud services, but I think your view might be a little skewed here. Does a senior mechanic need to understand the physics of piston design to be a great mechanic, or just gather years of experience fixing problems with the whole system that makes up the car?
I’m a Senior Systems engineer. I know very little about kernel programming or OS design, but i know how the packages and applications work together and where problems might arise in how they interact. Software Engineers might not know how or don’t want to spend time to set up the infrastructure to host their applications, so they rely on me to do it for them, or outsource my job to someone else’s computer.
Does a senior mechanic need to understand the physics of piston design to be a great mechanic
I would argue that if senior mechanic doesn’t understand the physics of piston design at least on some degree he’s not a great mechanic. Obviously mechanic doesn’t need understanding on metallurgy, CAD models and a ton of other deeper level stuff just like an IT engineer doesn’t need to know on a deep level how circuit boards are designed or how CPU die manufacturing process works. But both benefit greatly when they understand why something is built the way it is.
I’m also an systems engineer of sorts and have worked with software engineers. And I’ve had requests like “Can’t you just set 'bind-address = 0.0.0.0 on mysql-server and disable firewall” on a directly internet-facing machine and then received complaints when I’m “making things more difficult” from “senior software” -titles. Sure, I can’t write the code they’re doing, or at least it would take me a crapload of more time to do that but on the other hand there’s guys who have so very narrow understanding on anything they work with that it makes me wonder how they can do their work at all in the first place.
Of course no one can master everything in any field but I find it concerning that a lot of guys just press the buttons more or less randomly until their thing works without any clue on what they actually did and how it might affect on different parts of the house of cards they’re building.
But you know what the kernel is. You know that syscalls are a thing, you know what role the kernel performs, you know that different filesystems have different properties (and pros and cons), etc…
You don’t need to know the details, perhaps, but you can’t ignore the fundamental theoretical concepts of kernel and OS. You might not know the whole detail of the boot procedure, but if your machines are stuck on boot, you know at least what to look for.
Here I was talking about equally foundational topics. There is nothing “above” - say - producing attestations and then verifying them. That’s literally all there is to it, but if you don’t understand the theory behind it, what exactly are you doing? As as I said, I don’t care about the details, I didn’t expect someone mentioning ciphers or timestamp authorities, transparency logs etc. All I would expect is “we produce a signature with a bunch of metadata and we verify it where we consume the artifact, so that we are sure that the artifact has the properties attested by the signature”.
Not knowing this is like someone claiming that they administer Linux machines but can’t explain what network interfaces are or how routing is determined. This is not a question of being expert on different layers, this is just being oblivious to those other layers completely.
deleted by creator
