I also don't understand why care about result if you are working in a big company and you have fixed salary. Most of the time result isn't about better user experience, but about profits for shareholders. I guess all people are different

I think I am just not result driven. I like to do stuff myself, not providing info to some super intelligence to do cognitive work

This role is also require way less qualifications and cognitive ability. You need local knowledge, not fundamental knowledge: why in this exact company things the way it is. Mostly not transferable

But architecture creation can also be delegated to ai if you give agent enough context about functional/non-functional requirements, certain business side restrictions etc. Your work is to be feedback loop mechanism between people and llm, no? Please, give me encouraging examples of technical tasks that cant be delegated to ai with accurate prompting

It is just made my work unfun for me, that's it. I always loved problem solving, writing code, thinking about architecture. Now all of this is delegated to LLM's. Only things left is providing documentation and evaluate solutions. Evaluation part is present only because documentation/harness isn't perfect

Your are not building software together with Claude, you are only a context provider. You may think "ah, but my expertise is irreplaceable", that's false.

Another side effect is that code doesn't have any value. You can code new search engine from scratch and people will think "huh, I could prompt that too in a weekend"

I guess not many people liked writing code like I did (codeforces red btw, but this is also dead now), so effect of AI is net positive. But I give up. I need 2 more years to earn enough for the rest of my life, and I will never work in coding related field again

The irony is I have unlimited tokens, because my work is to create and deploy SOTA inference solutions for LLM's in one very well known provider. I've used all types of LLM's for a large range of problems for a long time, so this is not a hysterical post after discovering Claude for the first time

What is exactly your work? Give context to llm, review, update context. Navigating some sort of super intelligence thru your company's harness is not the same as writing code and creating ideas from scratch. And I don't understand what's fun in that

Yes, you can ship software faster, make corporation even more money. Why is this even important for regular worker? I liked the craft itself

Now I am just a monkey thats: 1) add enough context, description and harness to an agent 2) review ouput and repeat 1) if context is lacking

It was bottom to top: from understanding to implementation. You were the owner. Now it is top to bottom: get implementation first, try to get understanding later. Thinking is also delegated. "Think" nowadays is "reformulate,answer questions, add context, try again". This doest feels like I am doing the work, this feels like I am a limiting factor here

Another side effect is that any code now have 0 value. No one evaluating how you guided an agent, what decisions you took. People seeing your work and think "ye, I could vibe code that too with enough time" even if this is not true

And my work isn't css and html (with all respect). It is mostly high performance clusters, parallel computing, OS, low level, SOTA online llm inference etc

Now I am seriously considering blue collar job, as I have more joy building stuff with my hands than to be a passenger/context generator to an ai. I am not a business driven person, I don't really care how much money my company earns (sorry). I just like to solve technical puzzles and think hard

P.S. yes, there are corner cases ai can't do well: non trivial, highly specific algorithms and implementations; complex patches to gigantic multi domain proprietary code bases, but that's like 5% of my work