Long live the AI slop
Every time I close my eyes and open them, I feel like a new AI-tool/assistant/ultra-wow thing is staring at me! đ
It is unbelievable how many things are vibe-coded in 1-2 hours, they work while testing them, they are communicated as fully-fledged products and then they are left to their fate. Last week I had an experience that is worthy of a smile! At the company I work we have already hugely embraced the AI lifestyle. We speak AI, we feel AI and we try to leverage it as much as possible. Most of the times it seems to work. I mean we have automated many things, it has increased our productivity by an important amount and it seems that the experience can only get better. But on the way to success we have many attempts of creating things that were not successful! I wanna talk about it to cheer up!
Well, when work from the office, we have food from a catering service (which is really nice)! The procedure of ordering is as follows:
Each week until Wednesday at 1 a.m., we have to select the food that we want for each day of the next week! Every day has like 7 (on average) dishes! A few hours before 1 a.m. an e-mail is sent to us that reminds to choose food for the next week!
This flow seems a bit archaic, especially when we are talking about a tech company that really embraces the automation, but it has been serving as for quite some time, therefore nobody spends the required time to update it! The only thing that is a bit problematic is that it is quite often to miss the reminder on Wednesdays, which leads to weeks without food.. đ˘
Anyway, last week, while we were at the office we received a message from a new bot on slack (around 11 a.m.) that said something like âYou have selected to eat chicken with rice for todayâ! It seemed nice and everyone really liked it! We thought âfinally, we will stop using the excel to select our food, and it will help us stop forgetting about it!â. We made thoughts about how it will be scaled to automate the whole process etc.
It seemed like it worked only for that day, because from that single time that it worked, it havenât worked again!
But, what is the point of this post?
It comes together with the fact that many open source repositories have started to stop accepting pull requests. The AI slop is here to stay⌠It is sad but true that many people are producing a lot of stuff that is actually half-baked and eventually (in the long term) it creates a lot more issues than it actually solves.
Itâs a bit of an odd era, because we are not getting exactly what we were promised and everyone seems to not care about this. Again it is really important to mention that we were promised super intelligence and we are getting code-completing on steroids, which acts as super-intelligence. Last weekend, I was experimenting with small LLMs that can run on my laptop, like qwen3 (I would like to make a post about it) and it was incredible that the relationship between âhow wrong it is/how much it is hallucinatingâ and âhow sure it pretends it isâ was inversely proportional!
I believe that we need to start to think how we are going to make LLMs be more reasonable by them-selves about what they can and cannot do, because apparently we are not good at this! đ
