
@Pirifani
iAccount based inSweden!
About this account
- Account based in
- Sweden
- Connected via
- Sweden App Store
! X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Passionate about the art of crafting code, books, innovation and all the little things in between that lead us to become better engineers
Stockholm, Sweden
Joined July 2016
- Tweets227
- Following461
- Followers51
- Likes785
Alek Pirifan retweeted
how do i explain to my PM that i can't show any results yet because i spent the entire week making better architectural decisions, while my coworker Chad one shotted a comparable task in an hour and satisfied the business immediately?
i can't say "ask me in a year when you have to maintain Chad's code / when you see the CVEs", my performance review is today.
not implying anything here, it's a genuine question
Replying to @meowkoteeq
This, but unironically. You should disable as many features as you can bear. It will sharpen your mind, and force you to keep a more complete picture of your code, which leads you to make better architectural decisions.
Alek Pirifan retweeted
When it comes to writing w AI, there’s such a cognitive dissonance between the person writing vs those reading.
Those writing: “AI saves so much effort! And it’s much better vs what I’d have written! People will think I’m so articulated.”
Those reading: “WTH is this AI slop?!!”
(Some) writers appreciate LLM writing assistance but readers hate it. Survey says… writethatblog.substack.com/p…
If readers suspect AI, they not only stop reading, but also downvote/avoid the author.
Alek Pirifan retweeted
I’ve realized today that AI psychosis is much worse than I thought it was. It’s profoundly disturbing, and I’m worried it’s only going to get worse.
Alek Pirifan retweeted
inconvenient truth: generative AI code penalty. this apply to software and coding in particular as well. when you outsource the work to an AI, you don't learn anything and end up with worse product (on many metrics)
Alek Pirifan retweeted
Well this is damning and explains SO MUCH about why Meta is losing its best engineers; why morale is rock-bottom; and why an engineering culture built up in 20 years was destroyed overnight.
Zuck + execs wanted to lay off ~60% of staff.
When the business does better than ever.
Alek Pirifan retweeted
It's not unlikely that we will quickly get to a point where companies are hiring engineers solely to stop bad code from being merged.
Alek Pirifan retweeted
The new technical debt is AI-generated code committed by non–software engineers. The code gets merged into the code base and gradually convolutes the entire repository.
Eventually, the code base becomes a spaghetti of AI-generated slop that not even the best LLMs can organize and refactor. It just becomes an endless nonsensical stream of patches and dead code.
Alek Pirifan retweeted
What an embarrassingly lazy take. Perhaps some engineers are struggling because they *are* used to being outcome-oriented, and their jobs now have them pulling a slot machine handle 1000x a day, where more than half the time the slot machine spits out shit instead of money? 🤔
Alek Pirifan retweeted
Engineers who really care about the quality of their teams work are getting burnt out. Overwhelmed by PRs, difficulty in explaining why implementation choice maters, noticing that its a much easier policy to not care and slop it up. Wonder if it's a pattern..
Alek Pirifan retweeted
What I'm hearing from ~3 such stories:
Meta went too far firing people. They are now doing MASSIVE counter-offers like to those handing in resignations. Some accept it, most just use it for more $$ at New Job, desperate to get out.
It was such a predictable outcome btw:
Alek Pirifan retweeted
I let the agent iterate on the same bug 5-6 times. It kept “fixing” things endlessly.
Git history showed it actually solved the original problem just fine…
…but completely rewrote the implementation style every single time!
like a bipolar developer with zero consistency or taste
This is why pure agent loops without strong architectural ownership are dangerous.
Alek Pirifan retweeted
if prompt "are you sure?" in a loop, you never exit the loop and the product will eventually collapse. but the agent will be apologetically fix things until the end of the world.
Alek Pirifan retweeted
Totally agree with this. Every take I'm seeing on software engineering in the age of AI has serious holes in it:
- Let the AI agents write all the code and never look at code again? You will be held accountable when the code breaks down.
- You need to review all the code that the LLM produces? Not going to happen. There is just too much code and you'll end up possibly becoming slower than writing the code yourself.
- Engineers will just become managers/orchestrators of AI agents? Maybe, but AI coding agents are very different from human coders. They don't "learn" in the sense that humans do. They just adapt to the changes in the behavior of the person who manages them.
We are indeed in uncharted territory.
Alek Pirifan retweeted
This effect is really common in software. The Android dev community spent 10+ years doing something called “Clean Architecture”, just to end up realizing that most of it was not needed 🤷♂️
Folks would create 5+ layers of code that couldn’t touch each other for some reason, so they’d throw an interface between every pair of them, then force mapping pojos from one layer to another. All that just to build a simple app.
This effect is very hard to escape. The way it happens is:
- someone learns a new concept that sounds revolutionary in theory.
- People start writing and talking about it in complex terms as if it was rocket science, just to position themselves as an authority on the topic and grow their personal brand.
- A whole scene is created out of a concept that was meant to be a trivial thing, to the point that there is a major force pushing every dev in the community to do the same.
Usually only a handful of devs are smart/brave enough to study it with a critical eye and propose alternatives. Mostly because they are hardly pushed back by the whole community when they do so, since the big majority of devs simply follow what everyone else says, and never try to actually understand the concepts and why they exist.
Alek Pirifan retweeted
Vibe coders trying to justify why they don’t review the code and rely on their "sophisticated" test suites that are also vibe coded is the funniest shit
Alek Pirifan retweeted
A very annoying thing that happens now is that people who know nothing about a subject will "consult" AI, get the wrong information, and then try to "correct" you on a subject you actually know something about. They treat AI like the Oracle of Delphi
Alek Pirifan retweeted
The amount of AI writing on OpenAI's docs makes me sick.
Filler sentences for nothing. Mannerisms and phrases humans would not write. Has a human even read this?
And all of this will change how other docs are written, and how we all talk - for the worse IMO.
🤮
Alek Pirifan retweeted
here's the thing
when I work with peer-human signed code, I have trust in its abilities and I can even skip review. It's a matter of trust. matter of shared experience. and it works.
when I work with agent, every line of the code, every decision is a matter of suspicious activity. I have enough things to set this mindset, and went the journey that put me in this position.
agentic coding plateau. models are less capable of reasoning (never been actual thinking). if I want to continue use the tool I need to refrain from giving it a credit of trust, it does not deserve (because the tech behind it does not allow it).
this is over. the discussion about mythical capabilities is over.
Alek Pirifan retweeted
one of the biggest annoyances in working with agentic debugging/problem solving is how it is focused on fixing results and completely inability to find the core issue at the same time. It gonna build a pyramids of mitigations and see no "why".
this message comes from a frustration of last two days, where I encounter problems no agent, no model is able to solve properly, in two different projects, despite complete logs and source code available. And only after a day of prompting I sit on it with pen and paper, and actually dig with understanding the process fully.
I got lazy i many things. That's not what I was promised. It happens quasi randomly. it slows me down effectively.

























