I’m building an LLM app, I hate it
A few months ago, we did a course at work on the basics of LLM. My boss was thrilled.
The first thing that came to mind was “All the hype surrounding this sector is turning against me.”
I thought the bubble would burst and all the time would have been wasted, but I was wrong. We are still in the midst of the generative AI bubble.
Months later, I decided on my own initiative to create an LLM-based app to distribute to our customers (a few thousand) based on a generative model that uses our software documentation to answer questions.
But why?
I did it to take advantage of the current bubble, even though on an ideological level I hate the very idea of polluting in order to generate meaningful sentences.
I did it to remind my colleagues who are stuck 30 years in the past that they have no idea how new technologies work.
I did it to learn something new that could be sold to other companies. Ultimately, changing the topics of the rag changes the experiences of the model, so as the sole owner of the source, I can resell it.
I did it so I wouldn’t have to keep doing boring, repetitive tasks that waste my (limited) talent and my (moderate) experience. Change the query component yourselves, idiots.
I did it because some time ago a colleague attacked me, saying that I shouldn’t take initiative in the code structure, because otherwise the boss would get angry with him, who is also responsible for what I do.
I did it to feel competent in a programming language that I have mastered, of which they only know the name.
I did it to prove to my boss’s boss that he saw the light when he decided to hire me, but that he is wasting resources by making me do codemonkey work.
This rant is not about LLM.
What did I learn from this situation? Technology isn’t bad, and economic bubbles burst sooner or later. But I realized that you shouldn’t use tools built by others to build your own things. It’s much more useful to build your own tools, not to get a good product, but to improve yourself.
It doesn’t matter if the market asks you to build an AI-based app, or if you decide to do it yourself, the important thing is to be good at building it.