I basically neurodivergented myself into a well paying niche job. I live in fear of being re-organized into a proper job org where everyone just does one thing all day long. I would be like one of those sad animals at the circus in a little cage.
3 Likes
Iāve been trying so hard to change my scope so itās not just security review cause that has been slowly killing me. The one tiny specific function as a cog in a larger machine is so painful to me.
1 Like
A couple of interesting articlesā¦
1 Like
This is called a truck driver and it is mind numbing
2 Likes
I do not envy my network ops peers whoās only job is setting up and managing vpn all day every day.
3 Likes
Ngl, this is my goal (because of the same fear); job security through being uniquely indispensable. Preferably, Iād be more of a jack of all trades (because I gained all the ADHD symptoms but none of the perfectionism; āgood enoughā is my motto).
1 Like
Right? AI can code but can it also argue with vendors on the intricacies of VPN laws in the Middle East, diagnose voip issues for the Philippines, hard phone issues for India, network latency for Jamaica, update 80 legacy hosts to respond to a zero day, and make memes that make my boss laugh? I hope not because then I really am screwed.
1 Like
Now thatās where we differ I have no interest in being uniquely important in any way. Iāve made myself redundant so I can easily take time off making sure I have multiple people on my team who can do what I do.
My biggest goal is to be a personality that people canāt imagine leaving and a connection for my customers and leaders that is pivotal to the team through connections and knowledge., my job function can be done by anyone including AI some day and making myself pivotal wonāt protect me from the layoff hammer cause Iāve seen it first hand companies will not care and will hire 6 people to replace one critical person just to save a few dollar this current quarter. Might as well not stress myself in the mean time and create more work when I can be useful but take time when I need to without being on call.
3 Likes
Yeah itās a real fine line. My career success was basically I started as a holiday temp doing phone calls and every time someone handed me some special work I automated it. Replace this with an excel macro. Replace that with a python script. Somehow 16 years later Iām an SDE II. Iāve just spent 16 years replacing every job someone handed me with automation and getting a different job. Rinse and repeat. They could replace me with 4 engineers living in India for cheaper than me. Maybe some day they will. Like you said being indispensable often means a target on your back, but you also donāt want to be super super redundant lol because then you get hunger gamed in the layoffs. You need to be juuuust important enough that people miss your presence on vacation and stuff starts to fall apart a little.
2 Likes
If I had a personality outside of yoyos, video (games/movies/tv), or computers/IT, I would absolutely do that. Unfortunately, Iām not a social butterfly, a terrible actor, worse poker face, and small talk visibly angers me, and I donāt have the time, ability, or willingness to manage dozens or hundreds of various types of relationships (Masking is real, yāall). Granted it doesnāt help that my memory is basically just a hat full of sand, each grain engraved with the target memory and recalling is the equivalent of me using the grabbing machine game trying to get one specific grain of sand.
I also donāt care to lie or embellish myself, especially not to a bunch of strangers, even in the name of networking, for example, dev not engineer. Tangentially, I donāt make/stay friends with people where I feel the need to have do that, but I digress.
Would networking be easier? Sure, if youāre that type of person, but Iām not, so I cope in a different way, is my way foolproof? No, but neither is yours. At the end of the day, unless your on first-name basis with the person making the final call (or, I suppose, on first-name basis with someone whoās friends and has sway/pull with the guy making the final call), job security is only as strong as how much/little you cost.
All in all, for my method of cope this
is the linchpin that ties it all together (Personally, Iād replace and with because, but like I said, Iāve got the personality of grey, hard plastic cup that says āIām lazy and chaos incarnateā in clear typeface but is large enough to read without having to get too close
No greater short horror story than this
Oh, also, I like to talk, have no filter, and donāt know when to shut up
1 Like
This feels so right for this week where Iāve been forced into several meetings about agentic AI and how we arenāt vibe coding we are creating developer efficiencies with specialized prompts. (Aka vibe coding)
Also watching as we implement AI into our workflow for IRM and how my junior colleagues are literally excitedly testing this trash system thatās designed to replace them so that me and the senior engineers can do the job with less overheadā¦ they just donāt understand that last part yet.
I call it trash because itās an in house model thatās on the gov fed environment so it canāt access anything outside of our firm and also canāt train on existing data for compliance reasons meaning the AI is dumb as a brick and had to be trained by my junior colleagues using it and teaching it the right signals.
Currently an AI generated system security plan with the in house model for an azure app keeps trying to insert details about s3 buckets when its blob storage and mentions idrac for some reason in the physical boundaries which Iām trying hard to figure out why but canāt be bothered to be honest.
Iām pretty sure itās built on copilot too but I havenāt really cared to learn more about this systems inner workers cause not my monkeys even though they sure are running my circus
3 Likes
This guy seems to explain some of my fears about the AI bubble that I struggle to articulate myself.
Also the anthropic Super Bowl ads against open AI are hilarious
2 Likes
āGovt funded open AI bailoutā is the takeaway form this video. They gonna print. Thats whatās different from the 2000 bubble this time.
1 Like
Opus 4.6 is out now and itās 5 times more powerful than Opus 4.5. The models are on self learning loops and itās gone exponential now. Chat GPT agent mode can easily do stuff now that it wasnāt able to do for me just a month ago. Iām sorry developers, but if you arenāt learning how to control a team of agents youāre cooked.
Itās a problem nowā
1 Like
It is clear from this conversation that the experience between software developers and non-developers is quite different.
Developers:
- Create applications with specific needs
- Adhere to specified interface conventions
- Have to make specific changes to support specific features
- Implements custom behavior and custom business rules
Non-Developers
- Create an application that kind of meets their needs
- Those needs are ill-defined
- Security is a non-issue - since they do not understand the threat
- Specific business rules are unknown
- Testing is done by doing.
This points to different expectations for different groups. I can whip up a new application framework quickly using AI. Is is very hard to make AI do EXACTLY what you want it to do in code. Our AI is completely befuddled by our corporate style-guide for web-applications; api interfaces and pipelines. It creates applications is pure html and then I need to re-write it using our component library and custom CSS to implement UI standards. This saves me no time.
A non-developer does not really know what they want before starting. They morph AI code into something that they feel will meet their needs through trial and error. Corporate standards mean nothing; as long as it looks cool. This does not work as well when you have a specific solution that needs to be implemented with specific UI-behavior, implementing dynamic UI features based on custom business rules.
Our company is using AI-agents for corporate workflows. We use them extensively to triage patient intake as well as routing patient questions and answers through our doctors and admin by screening the questions and passing them to the correct person for reply. But these are not software applications. they are corporate-level, designed workflows for moving information within the organization. This is where I see the greatest threat to jobs.
I can see agent-workflows displacing many white-collar jobs that perform these agent-functions now. It may also replace many existing software applications that support this workflow; as well as replacing many applications that might otherwise be developed in the future as the agent workflow begins to replace multiple software applications being used throughout the organization. Agents, instead of people using applications, will move information into a string of agent-operations where the AI saves and utilizes relevant information for the workflow that might otherwise be stored in a traditional database application and accessed with traditional web applications by human operators. The result will be less people; using less software applications to perform the same function
The ultimate question will be: does AI replace people? Does it replace the traditional āapplicationā?
Or do AI-agents free up more people to do more things within an organization? Do layoffs lead to more people working elsewhere doing more productive work?
Have you tried Claude Code yet and ran multiple agents in parallel to develop software for your specific needs?
Yes, the problem is that AI cannot āseeā what it has created. It thinks it has done everything just right - but there is no proper use of CSS and the screens are all funky. It struggles with layout and is constantly trying to use its own CSS rather than our styled components when I ask it to fix certain visual aspects.
You tell Claude to make a table with certain columns. I cannot simply do that. Each column must be a certain width, have a certain sorting and it must have the correct colors, indentation, look and feel as defined to the pixel by our style guide. Give that a shot.
It definitely can āseeā what itās done. Just show it a screenshot of it. Show it a screen shot of what you want. Create a skill for your exact specifications.
Try this. Have one agent with a set of skills and .md files of your exact build specification. Have another agent that takes what agent one builds and validates the work by measuring everything against your company build specifics. Have it create a plan for any needed changes and give it back to agent 1. Put this in a loop until it gets it correct.