How to use LLMs (the correct way IMHO)⌗

Since when LLMs started to propagate in our lifes, I felt like I’d become useless one day, and that I’ll stop to learn what I was learning because there was no more need for me to do so.

And I’m sure that, if you’re reading this, probably you felt like this too. Maybe you still feel like this. Maybe you think that you’re useless in this society, and being skilled is something that… is not anymore required at all.

I found confirmation of this idea many times in many ways because of many people that didn’t know a fraction of what I’m capable of doing by my self and a little documentation, but still they bring up things I couldn’t write or think of even if I worked on it for days or months.

The feeling of being useless became normal. And so, what I once called “Experiments” became the way I did things but never admitting that I was, in fact, deskilling.

Which brought me to just stop thinking clearly and doing things with so much more trial and error than before, so much more blindly than ever and so much more frustrated of not being able to think better than this.

But at the same time, I didn’t want to became like some people that I know which are so lazy that instead of writing a simple docker system prune they prompted Gemini to do that many times, so much that a single docker system prune consumed their weekly token limit.

That’s another level of laziness but people like this are actually making more money than me. It is unbelievable that they are even making money in the first place, imo.

Ok but how do you use AI now?⌗

Well, wait, we need more context.

Firstly, I thought “ok I know I’m deskilling, I know I have a problem and I know this is actually driving me crazy”, and I mean it! Psychological effects on this matter are documented many times in literature, and I was experiencing it. Well.. I’m still there. I’m not “cured” by the “illness”. I’m literally just coping… better than before. Or at least I’m trying to.

So for some time I just decided to stop using AI.

It didn’t work.

So I decided to “ok then I’ll block AI”.

That didn’t work too.

So I decided to, honestly I don’t know what I was really thinking but for some reason I just said to myself “if you use it you are exactly like them” and this for some reason helped me.

But it didn’t work neither.

And finally, I think I got an answer!

Really, this is how I will use it from now on (and let you know if it works later)⌗

Don’t use it for things that you know how to do. If you know WHAT you want, and you know HOW TO, then just do it. Don’t lose time with LLMs.

  • 1: Don’t be lazy (don’t use AI because you are lazy)

Then, if you don’t know what you want, and you don’t know how to get it (somehow this makes sense to me), then:

  • 2: Don’t ask LLMs what you’re looking for

Instead:

  • 3: You can ask them to find relevant information for you

BUT:

  • 4: Don’t ask EVERYTHING. Keep every search as minimal as possibile. Fallback to LLM only if task is REALLY complex

And then I effectively realized one super important thing: maybe, just… maybe…

  • 5: ..I should use AI only when other ways are ACTUALLY not possibile

Remember that LLMs are a great technology, and using it for simple things like… writing emails?… is shameful.

Don’t be lazy, you should be the person that writes the email. Don’t be lazy, you should be the person that corrects grammar of you’re email (of course using tools for assisted correction is ok). Don’t be lazy, you should be the person who runs docker system prune

And please, don’t be lazy… if you actually need help… ask a human to help you.

See you on the next one.