What is the future? Does nobody knows?
Posted by ovidyel@reddit | sysadmin | View on Reddit | 58 comments
I’m hitting 42 soon and thinking about what makes a stable, interesting career for the next 20 years. I’ve spent the last 10 years primarily in Linux-based web server management—load balancers, AWS, and Kubernetes. I’m good with Terraform and Ansible, and I hold CKA, CKAD, and AWS Solutions Architect Associate certifications (did it mostly to learn and it helped). I’m not an expert in any single area, but I’m good across the stack. I genuinely enjoy learning or poking around—Istio, Cilium, observability tooling—even when there’s no immediate work application.
Here’s my concern: AI is already generating excellent Ansible playbooks and Terraform code. I don’t see the value in deep IaC expertise anymore when an LLM can handle that. I figure AI will eventually cover around 40% of my current job. That leaves design, architecture, and troubleshooting—work that requires human judgment. But the market doesn’t need many Solutions Architects, and I doubt companies will pay $150-200k for increasingly commoditized work. So where’s this heading? What’s the actual future for DevOps/Platform Engineers?
Junior_Option1176@reddit
I'm definitely more worried about the cloud than IA. I see cloud becoming insanely good at managing itself, autoscaling and simplifying infra to the point all you will have to do is throw app code at it , tick a few boxes and pay the bill. Now with cybersec becoming a big deal, bespoke solutions and artisanal configs will go away.too.Auditors will want every infrastructure to be standardized so companies can use insurance and things like that.
IA will only scare me when energy gets extremely cheap. Probably requiring nuclear fusion.. And something like AGI becomes possible + massive datacenters are built. By that I mean pretty much AGI where you won't need the concept of apps or software at all. By then pretty much every job is dead anyway.
rcp9ty@reddit
Everyone worried about AI replacing their job has forgotten the core word in their job title of "system admin"... Even if AI takes over coding scripting web page design whatever it's still requires a computer to run just be in charge of lots of servers every company is going to have a server farm or need someone to maintain their server farm. And the speed of which you can handle problems with a server farm dictates how much they pay you. Can The village idiot take care of a computer sure but the difference between you and The village idiot is that you might know all the parts in that server you might know how to replace them quickly. Or go a different route go with leadership training and be a CTO or IT manager or IT Director.
eman0821@reddit
I would be concerned if you think AI can handle IaC. You really have to be an expert to understand what the generated code is doing before blindly copying and pasting it into a production environment. You can take out an entire production environment with code you don't understand if it was never audited and maintain by a human. The code can be malicious, outdated security practices. Generative AI tools are designed to argument, not replace entire skill set or entire careers. It's a common misconception and big lie told by the media.
Subnetwork@reddit
Again, you’re looking at this very short sighted, it’s not what the technology is now, it’s what it will be, it’s going to keep advancing and getting better.
flurbol@reddit
Very good answer!
Just let me add: anyone who is copying untested code to production simply deserves the consequences... Doesn't matter if self written or done with a tool.
That said I am currently running a shit ton of AI generated code pieces practically everywhere in any system also in PROD. Never had an issue so far, but you wouldn't believe how much stuff I discovered prior to that in TEST and INT....
Subnetwork@reddit
I’ve noticed this, I don’t know if I’m just a more aware person, or I just think ahead more than the average Joe, but it always seems in any debate like this whether it’s tech or politics, people only look at the current snapshot and not ahead.
Myself as well, I even have the newer models evaluate the code older models have built just to see what it finds, each and every model is a improvement, I’m not worried overnight, but 5-10 years from now? Yeah I don’t see how everything isn’t going to be a lot different.
TopCheddar27@reddit
Then you would also know that MOST of the low hanging fruit for LLM optimization and training have been picked, and that process nodes are going at a snails pace.
Honestly at this point you are just fear mongering. An asteroid could hit us in 10 years and we're all out of a job.
eman0821@reddit
LLMs can barely do basic tasks as they lack crital thinking capabilities. Infact a computer doesn't think, it just only understands addition and subtraction done by the CPU. Agents are scripted tools written in Python that connects to LLMs to perform very very basic retinue tasks.
Subnetwork@reddit
Very basic tasks? You mean like most of the work you complete day to day as a sys admin and cloud engineer?
eman0821@reddit
Snall things like scheduling meetings and generating reports. It can NOT Triage incident response tickets, on-call duty when something goes down. It takes a human to understand that stuff esp as infrastructure gets more complex. You need to understand best security practices when provisioning infrastructure. No LLM tool can do any of that. Last but not least, you need an infrastructure for AI tools to run on. If the network goes do, so does the AI which is counterproductive if you ask me.
Subnetwork@reddit
If you’re only using it for scheduling meetings, then you’re not using it. I have Claude Code running 24/7 on a headless Ubuntu box I can remote into at any time, it can build and configure production APIs, even SSH to other systems and perform tasks, while constantly providing me feedback. For example I can say ssh using the key within documents and check for updates, reboot server, provide me feedback the entire time and verify services are running.
It does just that. Pretty crazy. I have a friend who has his own business with API related services, doesn’t even touch code anymore, has a Opus 4 sub. Makes thousands a month.
hell for M365 I had it create a auto pilot sync script and even custom window event viewer logs.
JoeyBonzo25@reddit
How are you keeping it on track while running independently? I agree with you and I'm fairly sure AI will be coming for my cloud engineer role before long, and I would like to be ahead of the curve.
Subnetwork@reddit
Try it with Claude code you’ll be surprised. Ask it to keep you updated.
eman0821@reddit
It's dangerous without supervision and lots of risk with cyber attacks, security exploitations with no human interaction. If you work in security, you should know the risk esp handling ssh and API keys. I don't recommend doing anything like that in a production environment. You can that in a homelab all you want.
And again, your agents wouldn't be able troubleshoot and Triage incident tickets or when a server or network goes down in the middle of the night. A human will always be needed in IT.
Subnetwork@reddit
I’m not talking about now, why is it so many people are soooooo short sighted and don’t have foresight? Give it 3-5 years and then come back.
eman0821@reddit
What do you think AI models runs on? It's still software that runs on a server that has to be maintained and scaled by IT profressionals. Once there is a network outage and that server goes down so does AI systems goes down. LLMs are written in Python that uses pytorch. MLOps Engineers which are essentially DevOps Engineers of ML deals a lot with the deployment of models.
Subnetwork@reddit
Research singularity.
eman0821@reddit
I disagree. You need an infrastructure for web applications and databases to run on. ChatGPT runs in the cloud on Azure in a Kubernetes cluster. IT roles has always evolved which is nothing new in IT long before LLMs existed. But so called AI replacing entire roles and industries is just smoke in the mirrors. You believe what you want to believe but it's all nothing lies been told. It's all a hype bubble at the end of he day. The bubble is already starting to burst.
Subnetwork@reddit
It’s called recursive improvement, it’s part of the singularity theory.
eman0821@reddit
I'm not seeing that as there's clear signs of slow downs in innovation. The AI bubble is getting close to bursting. It's basically the same thing as the dot com boom, bust era. Plus computer hardware architecture has major limitations. It doesn't take a computer scientist to understand that CPUs only understands addition and subtraction to compute binary machine code. A computer is worthless once you remove the RAM, storage or operating system. They are dumb machines when no operating system exist in order to tell the machine what to do. Computers have no thinking capabilities. They need code feed to the CPU to function. AGI is not technically possible because computers are far from replicating the complexity of a biological brain.
Subnetwork@reddit
You don’t need AGI to automate 90% of tasks done by everyday humans that sit behind a computer. A large part of even tech is taking one form of information and putting it in another form.
eman0821@reddit
You act like automation is new. That's not new in IT. Automation has been used for decades. I'm starting to think you are new to this field that doesn't have alot of experience. I've written scripts for years. When Puppet, Chef Ansible came years ago, that help do automation at a larger scale. Did it those tools replace IT roles? No. It's just a tool. AI tools is nothing different.
eman0821@reddit
Testing is one thing but do you understand what the code is doing? If you never took the time to learn how to code, I would be concerned because you can be operating up your infrastructure for all sorts of vulnerabilities and attacks.
Hegemonikon138@reddit
Same. I got rid of thousands of script files and snippets. I can now generate custom solutions on the fly.
That said, I know what I'm doing, what to ask for and how to test it properly, which is vital to actually leveraging these tools properly.
eman0821@reddit
It's all hype. It a bubble which is getting close to bursting as the evolution starts slowing down.
Nandulal@reddit
I think monday is coming
MajStealth@reddit
my last coworker always said he should have gone into woodworking, atleast you can build your own house and stuff for inside.
derekp7@reddit
The best thing to do is to find a stable company (a hard task as it is), with stable management / workforce (i.e,. where the entire IT department doesn't turn over every 3 years), and make yourself continuously valuable to everyone (your local team members, your own management tree, the managers in other departments, etc). Because people within management will constantly change, but you need enough continuity with others that have your back, that they can talk you up when someone new comes in.
And once they see how you are able to bring value even when new things come in, they will want to keep you around and want to always find a spot for you. That way you have some flexibility when it comes to learning what new comes down, and have a first crack at it.
Problem is, this takes a lot of work, and takes continuous effort, and can lead to burn out easily. To mitigate that, you need to shove as much money into investments as you can, so that you have a big enough buffer to retire early before you fully burn out.
buy_chocolate_bars@reddit
I heard goat farming is lucrative.
whetu@reddit
We have a TV show here in New Zealand called Country Calendar. It's been around since 1966 and focuses on rural stories.
So about a month ago I was cleaning up in the kitchen and had the TV on in the background, and it rolled over to Country Calendar. In this particular episode, they were focused on a small goat farm that produces artisan products like cheeses and so on.
I had just finished telling my wife about the running joke about leaving IT to run a goat farm, when the narrator of Country Calendar said
"... Mike, who works in IT..."
Oh how I laughed.
MaelstromFL@reddit
I actually have friends in western Colorado who run an Argo-Tourism business... With Goats!
They let people stay on their farm in an Airbnb form, and make goat milk, cheese and Apricot jams. They also have Sage production.
They didn't come from IT, though...
itishowitisanditbad@reddit
Whats it called?
Or like... how would I find something like that? Unless i'm an idiot i'm not finding it, its likely i'm an idiot though.
MaelstromFL@reddit
Sage View Ranch, you can just Google them.
Least_Gain5147@reddit
Here comes "goat herder AI"!
DeebsTundra@reddit
I reference that old thread at least once a month.
Jclj2005@reddit
Pig farming as well... need that bacon and pork chops
ConfidentDuck1@reddit
Put that on ice
ZobooMaf0o0@reddit
Somebody has to manage tech, a human will always be involved. The future is being an expert business/IT professional. What technology drives profits? Understand business and IT, you'll always have a job. AI can't make business decisions in the best interest of the company. AI gives you options but doesn't run the IT.
Expensive-Rhubarb267@reddit
It seems at the moment like the market is leaning heavily away from specialists & towards generalists. The ammount of role I see advertised for things like - networks administrator, cloud administrator, storage administrator have declined. Or if you do see those roles, they are broader than they used to be.
Most Sysamin jobs I see now involve some networking & IaC.
eman0821@reddit
Well Cloud Administrator was really never a real role that existed. The tttle was very niche to smaller companies. That job is done by Cloud Engineers. Network Administrator role died a long time ago as the operations and maintenance merged with the Network Engineer role. That's why Network Engineer title dominates over the traditional NetworkAdmin title in job searches.
Sysadmin role has evolved years ago since DevOps arrived but its really all the same stuff in rhe Cloud which isn't all that different from on-prem. Most sysadmins roles are Hybrid On-Prem/Cloud while others evolved to DevOps Engineer or Cloud Engineer roles.
Intrepid_Pear8883@reddit
I've had three jobs this year (2 contracts after leaving a full time). The knowledge needs to be general but you have to have a specialty.
Jobs now expect you to know a little about a lot, but a lot about a little.
The trick is to find what that little thing is they you want to deep dive on.
sdrawkcabineter@reddit
The future, is the merging of man and machine.
Intrepid_Pear8883@reddit
I still don't buy AI.
I'm in the corner of the current situation comes more from the cloud finally becoming mature. Most companies now are settled - so they don't need a lot of people shoveling coal. The last 15 years we saw VM consolidation then to the cloud. This is where we are now. AI is way too new even though it makes an excellent excuse.
It will change again. Like I'm in IAM. You really think companies are going to just give AI agents access to both IDP/SP? Hell no they aren't. So there will be work to figure out how to make it all work and secure. AI's impact will be in 5-10 years.
Foreign_Impress6535@reddit
I'm reminded of a line from Babylon 5, and it often seems fitting when you think about how fast-paced the IT world is and how quickly technology is replaced, and even how quickly WE are replaced.
“The future is all around us, waiting in moments of transition to be born in moments of revelation. No one knows the shape of the future or where it will take us. We know only that it is always born in pain.”
GhoastTypist@reddit
There's no such thing as a stable next 20 years. Every 5 years IT shifts to the next big thing. We're still in the AI bubble, once thats over the next bubble is going to be AI related as well but something completely new.
So lets just say this 20 years ago cloud was really just an emerging idea. The resources wasn't there to do it commercially. About 10 years ago cloud started to become the mainstream for software applications. 5 years ago machine learning started taking over, every application had some form of machine learning included. Now AI generative tools is the main trend and that was built on machine learning. So I'm sure in 5 years time we're going to be dealing with full AI suites where a company's backbone is AI. A system running in the background that ties all the infrastructure together, that all the AI applications talk back to. Think of it like a smart home but for the business.
In 20 years I'd be curious of if companies are anything other than top level "thinkers" in board rooms, and machines/AI interacting through websites or phone calls.
bulldg4life@reddit
An LLM can’t evaluate their code for being actually correct, can’t implement it in an existing brownfield environment, and it can’t refactor the code for thousands of companies running decade old software.
People saying AI will get rid of devops/cloud engineers always confuse me because I’m not sure if they are living in the same reality as me.
Will it most likely lead to less jobs because people will become more efficient? Yes, definitely.
But, you can’t sic an ai bot on a code base and no longer have devops or sre people managing infrastructure. And there’s still dozens of engg teams developing new apps all the time.
You should definitely be looking at how automation will change - I’m sure people will become ai engineers that implement an AI solution that does work. But someone still needs to maintain that and have it implemented in a production environment.
Subnetwork@reddit
You are ONLY looking at what the technology is now, at this current point of time. Give it a few years.
BuildAndByte@reddit
You need to learn how to use AI. You need to learn how to identify companies opportunities for automation. I have a Sysadmin on my team who is going to be let go if he doesn’t adapt… believes AI is the devil. Yet he takes half a day fine tuning powershell commands that ChatGPT can perfect in three exchanges.
eman0821@reddit
Doesn't mean ChatGPT is correct. You have to audit the code. I still write all of my code from scratch because i know it works and won't take down an entire production environment. You don't outsource everything to AI otherwise you are generating AI slop. I only use it to a small degree for repetitive boiler template stuff. ChatGPT doesn't replace a great programmer or scripter that knows what they are doing.
Doublestack00@reddit
*Yet
jrockmn@reddit
I use copilot for scripts all the time, but you still need to understand PowerShell to run those scripts. I’ve seen a simple mistake take down a datacenter. (Someone accidentally made a command “if LUN = in use then delete LUN” They intended to send the command “if LUN != in use then delete LUN)
peaceoutrich@reddit
That's not been my experience at all. LLMs can spit out pretty decent boilerplate and reduce toil and much of the actual "typing on keyboard" coding work, depending on how you prompt and use it.
I still have to do all the deep focus work to understand domains, context and all the moving parts. The coding part is only a minor part of IaC, at least in my book. It's the approach, the understanding of structured data and intents that makes a good IaC engineer.
Sure the tech will improve, I'm actually all for it. It makes me more efficient at my job, not redundant.
DayFinancial8206@reddit
LLMs will definitely reduce the need for many heads but realistically the way I see it shaping out is the people in the middle will probably need to stay. Entry level is threatened by it and high specialized code dominant roles are threatened by it. The workload will likely fall on the middle earners who now will need to leverage AI to do all the things they relied on people to do beyond that. It doesn't sound great for the industry.
However, and this is a big however, it is so far from perfect that we are going to need human intervention equal to the amount of AI doing work to "fix" a lot of what it is doing. When a new need arises, the competitive salaries for people who actually understand the code, systems and processes will likely increase assuming they don't need as many bodies. The thought of AI like gemini creating code for billion dollar orgs terrifies me.
BrainWaveCC@reddit
No one knows the answer to that question until they have access to time travel.
That's not how careers work. You try and make decisions for the next 1-2 years, and adjust as life changes with and around you.
There's no way to reasonably predict all the things that will be going on in the early 2030s, and how they will affect employment options.
SAugsburger@reddit
This. A LOT has changed in the last 10 years nevermind the last 20. Guessing exactly where we will be in 20 can be tough. There are some vague generalizations you could guess about the next 10, but a getting into details that would be very actionable might be tough.
Mustard_Popsicles@reddit
I learned to stop worrying and just take it a day at a time. No one knows the future, so it’s best to not stress over it. As much as I don’t like AI, it’s a tool that I’ve learned to used to cut down time, if it’s here to stay, might as well learn to use it properly.
LilRee12@reddit
I see most of the positions transitioning into these “overseer” types of roles. You’ll be covering more tech in general in your day to day, but not “directly”. You’ll need to have knowledge in a lot of different areas and will need to be able to use the AI to work on your behalf, but when something comes up you’ll have to be knowledgeable enough to go in and clean up behind those AIs every now and then. That’s where your years of knowledge would hopefully come in.
itiscodeman@reddit
People