#careerchange — Public Fediverse posts

https://mastodon.social/@h4ckernews/116556568722213579

https://www.seangoedecke.com/software-engineering-may-no-longer-be-a-lifetime-career/

My fuc..... response.

Firstly, comparing the arduousness of manual labor to that of intellectual work makes no sense. I can understand the overlap in principle, but not in consequence. For a start, the secondary sector is much more Taylorized (Fordized if you want), then the living conditions are much lower. And finally, the activity is much more arduous for the body and the mind than in the tertiary or quaternary sector (software eng.).

Talking about the loss of theoretical and practical knowledge regarding the use of AI or not is also something I would nuance. Yes, he or she who no longer codes loses in skills, but not so much in knowledge. Reflexes change, review management changes. The coordination of a project changes. But to go from there to saying that the person who no longer codes loses in knowledge is a shortcut which, in addition to being limiting, is fallacious. The engineer who no longer codes has issues with reflexes. In no case issues with understanding the code. And that is where I would put a nuance. The eng. always knows where to look for information, build their project, structure it.

From a practical point of view, they will write fewer lines, but in exchange, they will allow for better planning. They will certainly be less up to date on the use of a function, but they will be able to explain how the function must be encapsulated, and everything relating to micro-services or monoliths. In no way should AI make a decision. If you let it do so, you lose everything and gain an incommensurable technical debt.
Defend yourselves !!!! Explain to the paper-pushers that their AI is not going to succeed in explaining why such a technology is better for their project. The engineer or the architect will take everything into account, from OOP to the ultimate spec lost in the very depths of the JAVA doc regarding the Floating-Point Remainder Operator and why it is important. Calculate the cost of a bad operator choice in 5 years and tell your paper-pusher: "Do you still want the AI to manage your project in OCaml?"

AI is useful, I am not saying the contrary, but there is clearly a fundamental difference between a tool that makes decisions that have consequences and a secondary sector worker who uses an excavator instead of a shovel... (in both cases, he works in the cold, his pay is the same, but he kills his back less). He remains the master of his actions... Whereas the AI... there is a fabulation of domination and power and a dramatic misunderstanding. Who is responsible for the choice? The AI will never substitute itself for the responsible person (it's not me Madam Judge, it's gpt 8 that didn't pay attention that passwords must be hashed in Argon2 and not in SHA-1... it was in its .md though...)

So no, software engineers and architects will not be replaced, unless the statistical paper-pushers who calculate in lines of code spawned per hour assume the service interruptions, the maintenance and production release costs.

Today, everyone swears only by Claude Code and other "magical" crayfish supposed to do "everything" in our place. The result? They generate with pleasure all the bullshit that maintainers are desperately trying to protect themselves from: obese and incomprehensible PRs, impossible to reproduce bug reports, and feature additions that outright break the API because the tool mixes up terminologies without understanding the business domain of the project. I've seen AI proposals that didn't even respect the Single Responsibility Principle (SRP)... the basics!
So yes, this machine swallows docs by the kilometer. It spits out text with an incredible and fascinating aplomb (like the sexist boss who wants to make believe he knows your job). But it is plausible, never exact. And that is the whole difference with an eng.: instead of coding blindly, the human analyzes the system, the dependencies, the architecture, the production structure... the specificities... then finally decides, and makes the architectural decision, before delegating to the AI the drafting of the Slack message to explain to the team why we are not going to import such a bloated library just for the three features we need.

Pisses me off in the end :D

#HackerNews
#softwareengineering
#careerchange
#techindustry
#futureofwork
#softwareengineering
#Tech
#AIHype
#Architecture
#DevLife