Kiwibros with computer/programming/data oriented jobs. How secure are you feeling about your position? Are you making any contingency plans?

[grump]

(I know there are other AI threads but those are more general oriented. I am curious about the experiences of people, particularly those in jobs/fields currently most in the crosshairs of attempts at AI automation)


Those of you with white collar jobs particularly computer and data based jobs how do you currently feel about the AIpocalypse? Has it affected you yet or do you think it will eventually? What are you doing about it if anything?

Me? it is a bit of a kick in the nuts that one of the primary skills I've spent years cultivating might have been suddenly rendered useless for direct application. And I'm one of the people who enjoyed directly coding and having a hard time wrapping my head around especially the 'AI first approach' where you just let the AI do its thing and come in afterward if at all. And that to really get into it requires spending loads of money constantly on credits just leaves a bad taste in my mouth.

I am still of the opinion at the very least coding skills will remain useful for higher level applications of understanding, reviewing, and structuring projects. And that software engineers will still be needed for higher level expertise.

But still all the doomerism and blackpill I've been hearing on social media about how nothing is safe has made me a little anxious. Post upon post of supposed completely autonomous ph.d level agent swarms and crackhead high school dropouts who barely know how to use microsoft word supposedly making their own tech companies and coding the next Linux killer in 2 seconds. Of course a lot/almost all of it probably is astroturfing by companies and self promotion. But the mood does seems to have changed even within the past few months. Less than a year ago direct coding was still considered viable. Six months ago coding with AI assistance was considered viable. Now allata peepul/bots? are claiming you have to switch to an AI front and center approach. Practically vibe coding in all but name. Where it does virtually everything and you just double check. I've also had to repeatedly revise my personal projects that I am working on because it keeps becoming obsolete.

OTOH At the same time things seem to be changing so quickly they're seemingly not changing. Despite AI 10 billion xing everything allegedly we aren't yet seeing any apparent speedup on progress towards flying cars or colonies on saturn or a cure for all diseases. AI just seems to be this weird thing that somehow just vomits out slop art and makes getting and keeping a job harder so far.

But the scenario of AI completely replacing programmers and data scientists and analysts isn't even whats most annoying to me although its scary too (although depending on the exact scenario also potentially interesting to futurist me). Whats most annoying is the uncertainty and not being sure where to go and what preparations and skills to learn if any. Stay the course, expand your skillset, jump to a career you think is AIproof or AI resistant, it can all blow up in your face. And then theres all the chaos that will result from midwit HR and CEOs who are guaranteed to go the wrong direction from the one we actually will head. At least if AI takes over everything, which I don't believe it will but we will see, theres the possibility of not having to work anymore.

 

[estacks]

I decided to quiet quit 3 years ago and collect severance after seeing the writing on the wall: AI sabotages everything it touches. Once the delusional mania around AI pops your inbox is going to be filled with recruiters. There's no speedup of anything because AI makes even the best developers at least 20% slower dealing with all the slop errors. You spend 2 hours auditing boilerplate files that take 2 seconds to spin up with a schematic CLI. All LLMs can generate is a Ganges River of jeetslop from Stack Overflow posts and fail quality public Github repos. LLMs never generate anything original, which is 95% of the actual value of any tech employee. Every company replacing their talent with JeetPT is getting burned by multi-million dollar security holes and constant outages. If you think you'll enjoy auditing and cleaning absolute shit code for years of your career then keep on trucking and wait out the frenzy.

 

[Fcret]

I took severance and went on vacation for a few months. I already have recruiters trying to grab me but I'm enjoying vacation too much.

On the topic of AI and coding though, even the good shit right now is incredibly limited. If you're a programmer used to looking up shit on stack overflow, it seems magical, but normies trying to use it don't even have the vocabulary necessary to properly engage with a code generating LLM. Even worse than that, most of them start choking once you hit around 2 kloc and introducing tons of random noise into the mix so their ability to actually work on/debug even small projects is questionable.

This is why you see so many "AI IMPLEMENTED A TODO APP IN SECONDS, SWES ARE OVER" posts because anything more complex starts to really outpace what the LLM can actually work on.

Biggest problem right now though is that the best models are expensive to run and pretty much every company in this space is losing money just on inference alone (and then when you add in the cost to build out infra and train, it's a wash). It's also quite hard to cheaply add in additional data once a model is trained.

I don't know how far we can actually scale this technology. There are true believers and then there are doomers. If the last few iterations are any indication though, I think we're rapidly approaching some kind of fundamental limit on how effectively information can propagate through giant models (which is why the big new meme is multi-agent systems).

 

[ita]

I’m afraid it’s too late for me already. The place I worked at for years shuttered entirely a while back and I’ve had no luck securing a job since then. I’ve only been up to a mid-level individual contributor, so no official senior or managerial experience in this field to give me an advantage either. It has, admittedly, put me in a dark place lately.

 

[Absolutego]

Just from seeing how it's been rolled out in corporate America, even as bullish as I am on AI as a technology, I'm pretty bearish on its long-term effects on the overall industry. It'll be another tool in one's arsenal, and conceptual understanding of software is going to matter a lot more than syntax minutiae for people with a "developer" title going forward, sure, but that might actually improve the experience for the rest of us in the near term, since that makes Indians a great degree more useless to employers. Sucks balls for new grads though, since I feel like compsci degrees just train you on minutiae and code monkey jobs are the primary entry level position for the industry.

A full-on, super-minimalist AI implementation that requires maybe one or two architects to do the work a team of 200+ used to do is beyond corporate America's ability to set up (barring some rare exceptions) without non-tech managers fucking it up. More realistic implementations where you're forced to debug AI code are more likely going to be what becomes the new norm.

At least, that's what I've noticed being adjacent to it. As a backend engineer/systems guy/digital plumber more than a full-on code monkey, my day to day use of AI is just making convenience scripts take me 1/5th the time to write up, being free of errors from me not working in this particular app's preferred language for a few weeks and crossing wires re: syntax. It's also a pretty useful tool for asking dumb sounding board questions when troubleshooting without pestering coworkers, or rapidly evaluating a new tool.

 

[Fcret]

A full-on, super-minimalist AI implementation that requires maybe one or two architects to do the work a team of 200+ used to do is beyond corporate America's ability to set up (barring some rare exceptions) without non-tech managers fucking it up. More realistic implementations where you're forced to debug AI code are more likely going to be what becomes the new norm.

I mean you're going to need a handful of engineers for each product anyway. When shit hits the fan and you're losing millions of dollars per hour, you need adults in the room who can authoritatively fix shit instead of waiting for the statistical sludge machine to churn out the correct statistical sludge. Startups that "move fast and break things" (which are where a lot of the homegrown AI hype is coming from) don't really understand the value of each 9.

The real danger of AI though isn't to normal employees. We've dealt with downturns before and anyone who's been in this industry long enough should have realized how overinflated things were back in the early 2020s and started stockpiling their acorns for the winter. I'm more worried about infrastructure as a whole because now we have gangs of indians dumping the worst fucking AI code imaginable into all of our critical systems.

 

[Smart Turtle]

AI is fundamentally a tool, it doesn’t possess true critical thinking, intuition, or creativity. Its effectiveness depends entirely on the knowledge, skill, and judgment of the person using it. Someone who lacks expertise or understanding won’t be able to leverage AI effectively, It will take or ruin Jobs and Will create Jobs but the impact is over-exaggerated, You Will still be needed but it will make it much worse for medium to low level Jobs.

It will disrupt industries, particularly affecting medium to low-level positions that involve repetitive or rule-based work, making competition tougher and increasing pressure on these roles.

 

[Angry Alt Right Nerd]

From my experience with software engineers there's a pretty large part of me that thinks they deserve to be automated out of their jobs.

 

[Absolutego]

The real danger of AI though isn't to normal employees. We've dealt with downturns before and anyone who's been in this industry long enough should have realized how overinflated things were back in the early 2020s and started stockpiling their acorns for the winter. I'm more worried about infrastructure as a whole because now we have gangs of indians dumping the worst fucking AI code imaginable into all of our critical systems.

Yeah but that's a big part of why I'm bearish on the tech. Infra falling apart was a big part of why corporate america started onshoring in the first place (before cloud infra obviated the need to offshore at all), and just looking at the bullshit Microsoft is dealing with as the Jeets running Redmond turned AI loose on Windows 11 tells me it's not long before Azure starts similarly falling down, AI is made the scapegoat like it's the Dalit of all Dalits, and the industry as a whole does whatever the AI equivalent of onshoring is, which I guessed involves software development turning into a full-time debugging job.

 

[The Early 2000s]

Company keeps talking about acting upon integrating AI into what we do, and i recently had to take a long fucking course about how to use the company's own AI model to do almost everything. They had a section about how to have it create responses to emails from your coworkers. I swear if I see one of those em dashes in an email...

I use AI outside of the code editor as a better search engine tool and to help brainstorm some solutions when working. It's very helpful - beyond that, the amount of mistakes I've seen it make with more complicated issues has given me a piece of mind about ever being replaced by one. Not to mention that actually coding is only a part of the role. AI isn't all that great with collaboration.

 

[Tachibana]

My worries aren't with AI, but how the managerial class manages to completely fail people with technical skills. You spend months grinding certifications and shitty leetcode to do interviews, and then you get on the actual job:

• Everything needs to be shipped three days ago. There's inherent stress to the job.
• Having to deal with jeets more and more.
• Spending days in meetings instead of applying what you learned
• Multiple layers of management designed in such a way to insulate everyone from responsibility, but somehow fuck you over in particular if you get laid off

There is an expectation from this industry to grind on off-hours, because you are required to get past HR checklists and leetcode tests to get the job. That is starting to be the case for jobs that don't pay that well.

In recent years, the jeet invasion of tech has amplified. Businesses chasing profits above all else has led them to hiring either consulting firms based out of India, or just hire straight out of India. You, the worker, are expected to cooperate with these shit eaters. The Indian managerial class is by far the worst. They always overpromise, harass the shit out of workers to put in more work, and it leads to lacking architecture. When you end up in a team that's half or more jeets, you are in a "us versus them" situation. I don't see a solution to this in the short term as tech firms have named pajeet CEOs in masse.

You see this phenomenon too with services. It used to be the case that you would know how to program in a handful of programming languages. Nowadays, you need your AWS/GCP/Azure certifications, on top of knowing a bunch of subtools that are not "core" to anything and that you might not even use on the job at all.

AI is the latest scapegoat, I don't think it's as amazing as people make it out to be. Every project I've touched where there was some AI integration was a shitshow, and tends to be used as a bullet point to show a manager that you've contributed to the "company mission", even though there's no dollar value associated to it.

I don't think this industry is viable in the long term. I think one of the realities is that there are way too many hires who don't produce much value, and the personal toll you need to deal with to just keep your place is out of whack with the incentives.

 

[wtfNeedSignUp]

Pretty secure, AI just can't do algorithm design that's not textbook or ripped wholesale from stack overflow, so it's unoptimized as shit. Nevermind work on any project that's missing a large codebase to train on. And it absolutely shits itself once you have too many files.

The main benefit of AI is allowing fast testing of ideas. Lots of shit that would have taken days to write in the hope of doing an improvement can now be done in an hour.

AI is more of an issue to the managerial class than programmers. Even with the best AI has to offer, a manager wouldn't know the first thing about how to even compile a program, but a developer knows how to write an email.

 

[beyondtheclouds]

ITT: Motivated thinking

 

[General Zigger]

These opinions of mine come from the idea that I wish it worked as well as the AI bros said it did. I wish I had a super powerful 10x software engineer I could give good specs to and get good quality product out of. So many of my ideas would become possible overnight and reasonable to finish. Sadly I don't live in venture capital fantasy land.

Unfortunately, in my own (non-work) projects I work on things that either have little documentation, or the LLM models have out of date documentation. So they are limited to just being stack overflow replacers mostly. If you can get them to write working code, it comes out as a basic project that isn't ordered in any way for me the human to take over. LLMs routinely do not follow best practices for project organization. If I happen to get lucky and get a working project out of the model, the model eventually fails at the jump from low to intermediate complexity. If the project I am working on is too similar to something that has existed before, all the little details of that previously existing project enter the context of the LLM, polluting the project entirely. For example, I wanted to write an octtree library for a game engine, and the model consistently kept referencing Minecraft, even when my project had extremely different goals from Minecraft. This is most likely due to everyone and their dog writing a Minecraft clone and making it the biggest link between octtrees and games. I eventually just wrote it on my own to avoid fighting the LLM, but it did give me a good starting point to avoid the boilerplate.

For basic questions though it has basically replaced all internet searches. For now at least I can get a roughly correct answer 80% of the time without wading through SEO spam.

For my work, I work in an area where the tools routinely do not follow what their own documentation states. Error codes are in Chinese moon runes. The most detailed specs I get from my boss are the equivalent of essentially "it doesn't do the thing, make it do the thing, quickly please" and I have to hunt down what they actually want. Current SOTA LLMs do not work well in these conditions at all, and thats before we even consider that for legal reasons we are told not to use them because of IP concerns (even though the army of jeets ignore this).

I think that as a tool it has some uses, and it will remain after the whole bubble pops. But I doubt it will replace a lot of people. For that you have to worry about indians. Companies in my field/area are still hiring graduates just the same as they were previously. Notably these are not the big tech companies however.

(If someone older than I could cross check what I am about to say, I would appreciate it.)
I think what is happening is that tech has had extreme growth since its inception (around the commodification of the personal computer maybe?) and it seemed like there was a new frontier to exploit every five years. I can't speak on the past but I remember growing up seeing the change from software you installed on a computer, to software as a service. How mobile phones turned into pocket computers, with their own markets for apps and ads. My own relatives still talk about the change from calculating numbers by hand to using a calculator, to having the computer program do all the calculations itself. It seemed like there were infinite problems to solve. But after the rise of SaaS there was a lack of real problems to solve. This in turn leads to these massive companies desperately searching for any new avenue to grow into. So now upon seeing AI turn into a new hot thing, they are praying it works out so they can keep their stock valuations high, and at the same time are offloading costs by offshoring and replacing workers with H1Bs. Essentially I think we are seeing the contraction of the industry as the easy problems have all been solved and technology change has started stagnating.

I do think there are still problems to solve, and still value in writing software/designing hardware, but the problems that still exist don't have enough growth/revenue potential for the massive companies. Maybe we will see smaller and more focused companies as a result?

 

[The Hardest R]

I’m in management and I get pressure from CXOs to implement AI so we can start cutting jobs. That’s all they want. They think AI is an easy button that can make entire departments redundant because they got salespitched by a goon who claims a few enterprise licenses can eliminate millions of dollars of analysts and engineers. They will caveat it by saying we can’t take on any risk but at the other side of their mouth they want cost savings NOW NOW NOW.

There’s talk about companies learning their lesson already but it’s a long ways ahead, not when stock prices can shoot up now. Now is all that matters.

 

[Sooty Soot]

It's ok, I'm Indian and I'm basically guaranteed a six-figure salary no matter what, AI just makes my job easier.

 

[estacks]

My worries aren't with AI, but how the managerial class manages to completely fail people with technical skills. You spend months grinding certifications and shitty leetcode to do interviews, and then you get on the actual job:

Time for my real perspective:

It's almost like the tech industry is a concentration camp of all the worst people beyond all imagination trying to consume humanity's soul. Being rejected and ostracized by this system of doomscripters is one of the greatest acts of providence one could receive. Take the L (actually a HUGE W) and go live on a farm somewhere because the alternative for yourself is so, so, much worse.

The footsoldiers of tech, the ones that actually built this Moloch, bear more moral culpability for the absolute darkening of the human spirit taking place than any executive could ever claim. The executives follow their idiot beast greed to extract resources for vampires, but tech workers exert their full human agency to serve the destruction of humanity.

1. The gatekeeping is to keep out anyone who doesn't already worship arbitrary dominance hierarchies and pointless aesthetic knowledge (you will unquestionably sacrifice yourself to serve them for no real reasons)
2. HR commissars stalk you at all times to make sure that your personality matrix is ideal for service to power
3. Pay grades are tuned so that tech workers feel both proud of how hard they can flex on all the proles yet insecure at the fact that the C-suite makes 100x as much money (fostering pride and ambition to serve)
4. The shitty quality of life is tuned to extract maximal mental energy so that you're zombie walking too hard to understand the implications of what you are building (a transmutational system to infect all of humanity with Satanic thoughtforms derived by actual occultists)
5. There is a huge culture of psychedelics, amphetamines, and various other nootropics to "brain hack" one's own neural net into a mercurial slop that will never crystalize into any form of moral code or consistent principles (tech workers are the first in line to be consumed into non-agency)
6. Once they believe they have consumed the full extent of your generative abilities into LLM training data they will immediately fire you and blacklist you in the centralized HRIS platforms
7. All of these systems of entrapment are carefully designed to require your enthusiastic consent to participate in (you will always be blamed for the consequences of sustaining this system and you 100% deserve it)
8. It is not the act of witnessing or experiencing these disastrous systems that curses you, but letting them speak for you and flow through you until the point you have the Moloch AI systems writing your emails to your family for you. At that point, what are you? You are a tortured speck of a person who has chosen aesthetics over reality, Luciferianism over everything else. Now you are the bot and the bot will pilot your brain like a puppet.

If you think baring your soul and jumping through 6969 Leetcode tests to work in poverty among the most shallow, narcissistic, empty, trend-following nigger cattle is a great idea then you will get what's coming to you. If you see all the mind-jacked zombies doomscrolling so hard on their phones they kill themselves driving, and you choose to enhance this ouroboros of damnation, then you most assuredly damn yourself. We're well beyond the illusion that the tech companies are doing anything for humanity's good. We're already at the point where in the fringes of Reddit, 4chan, et. all you can find occultists who have fully replaced their human neural nets with demonic beasts of non-physical nature, thus giving those beasts physical nature. AI is the neural amplifier that accelerates this process, and just wait for the horrors that will take place as the darkness AI is comprised of is given robotic forms to attack us with. Whatever it will be, I assure you it will not be in ways you expect.

 

[Rungle]

Until the older generation completely dies out I am good with my cushy skilled service desk job.

 

[Ryo Ma]

Autism

Dude I just sit and tell the electric rock to make MVC forms, what the fuck is this