Fabio Manganiello on Nostr: I am an engineer with more than two decades of experience, and I use LLMs to write a ...

I am an engineer with more than two decades of experience, and I use LLMs to write a lot of my code in my corporate job.

I tried to first avoid them altogether and then minimize the usage, but I honestly don’t believe that an engineer in 2025 is given much choice.

You see, 20 years ago when I joined the industry (and still 10 years ago) it was possible (and actually it was the norm) to start a career doing C/C++ development, and then use the same languages and technologies until the end of your career.

The biggest leap you were expected to take at some point was how to use Boost or Qt, or a different compiler, but that was it?

Today, instead?

Well, in my current job alone I’ve already counted 7-8 programming languages that I’ve had to work on just in the past 2 years.

I’m expected to easily switch from a Java component, to one in Kotlin, to a Python script, to some legacy Perl stuff, to a Go executable that a guy made a few years ago, to investigating API calls in our frontend code in Typescript+React, to writing Painless scripts for ElasticSearch, Groovy scripts for Neptune, Scala code for Spark, Lua integrations for our C bindings etc.

And around 20-30 libraries and framework that I’ve had to master.

I’m expected to know everything about Airflow, DBT, Spark, Tableau, MySQL, Gitlab CI, Harness, Kafka, Redis, ElasticSearch, Neptune, Tinkerpop, DynamoDB, MongoDB, and countless other frameworks and data solutions. And I’m also expected to train on the fly my models with PyTorch or Tensorflow if required. And I need to know how to deploy it on the fly to our infra with Terraform, Ansible and everything in between. And I need to know everything about ingress, outgress, sizing and scaling cloud components, security etc. Oh, and let’s not even get started with the frontend madness, where the whole stack basically changes once a week.

It was just devs at the beginning, then it becomes full-stack devs, then full-stack devops, then dev-sec-ops, then dev-sec-ML-ops, now it’s basically “pay that single guy to know everything about everything, and threaten him to fire him if he bitches”.

At the beginning I was genuinely curious of learning new stuff.

Now, when the “new stuff” comes at the pace of one new thing every 2-3 days, and I’m given one day to master it, and by the next day I’m already supposed to deploy iron-strong stuff in prod with that stuff, I honestly don’t give a single fuck anymore.

No single human brain can hold so much information and master all of it.

And, at some point, it becomes even pointless learning so many things. You become the guy who already knows 60 languages and he’s tasked to urgently learn the 61st in order to keep his job, with no respect for the breadth of knowledge that you’ve already achieved over all those years.

So do you want me to learn Scala in one day, as if all the stuff I already know wasn’t enough for you, and the next day deploy a Spark plugin that needs to handle millions of requests per hour, otherwise you may lay me off, and you don’t even have enough passion in tech to listen to me and understand the challenges?

Sure, I’ll ask ChatGPT to make it for you.

If you ask idiotic questions and have unrealistic expectations of what a human can do, no matter their experience, then you deserve some vibe coding shit in prod.