I, mistakenly, have never given the entire AI trends that much consideration and even at one point suggested that it might be one of those overhyped technology trends that will fade away with time (like AR, Bitcoin, etc), . This post is about some of my personal experiences that made me reconsider this. This post is NOT written with LLM, though, lol.

University Years

My first ever experience actually doing something AI related, was during my studies around 2007-8, in fact, I did have quite a few AI courses at the university, including building a simple NN framework and even visualizing its internal layers structure and learning process with backpropagation. I didn’t give it much thought back then. It seemed to be quite a niche technology and just part of studies. I could see it classified some data I fed it with and saw how my classmate used it to recognize numbers from car number plates.

First Jobs

After that, for quite a long time, it hasn’t really shown on my radar that much. I guess I might have inadvertently used some tools that were utilizing some ML algorithms, like I remember using some open source library for a prototype project to match images. At Amazon I knew some teams that were working on AI related things, such as brand protection, recommendations, but never really worked on anything AI myself. Google is known to have been on the forefront of using AI long before anyone else. Advertising at Google has been running ML models for a very long time and I had to support their efforts by working on an experimentation platform that allowed those teams to verify their hypothesis and slowly roll out new models to the world.

Rasing Trend

During my years at Google, AI has risen in its popularity. Big tech companies started to invest extremely heavily into AI (Pichai-AI meme), oftentimes pushing for efficiency and cost cutting at one end, and at the same time expanding operations on the other end. I think it was also the time when I started making use of GenAI a lot more. More and more tools started to be available, coding has become somewhat easier, doing some summarizations has become easier and so on. Since I moved to META early 2025, the entire AI trend continues, and it’s clear that META is very aggressive in hiring top AI talent (media coverage). There is more and more interaction with ML at work, some of the projects my team is driving are to integrate with ML platforms, etc.

Crossing Personal Mental Threshold of Usefulness

My main reservation with using LLMs was that I usually felt the quality of results it produced did not justify the effort I put into prompting it. Especially given I always had to correct the result. Though, I believe this changed.

Last week I had to write a few new classes in C++, that would evaluate some expressions from configs, so instead of adding files manually, I just talked to AI, “hey, create me this and that and make sure interface has this signature”, “hey, add a UT class”, “hey, update dependencies”, and then it actually did a very fairly good job at all of that, not perfect, but really good enough to save me time. This is when realization came to me, this is now crossing that personal threshold I had in mind. It is more useful and worth a bit of effort fighting with it.

On a more personal front, recently, I wanted to replan some of my life goals, learning strategies, so I made heavy use of GPT and it’s just astonishing how good it has become at reasoning, structuring things, and actually producing what I want. I’m now a paid subscriber of GPT and am trying to use it more like a true personal assistant. I did use it before for financial advise, travel planning, summarization, etc, etc.

Last night, I was like, how about I ask GPT to learn something together, so I asked, “let’s create an AI learning plan, here is my background: …., make it personalized”. “AI refresher” was on the first week with suggested deliverables of building a small convolutional neural network on the CIFAR-10 data set. So… drum-roll, I asked it to build a notebook with code for all of it and it produced a bunch of code, which I followed up pasting into Colab, training the model and verifying its accuracy. It is just mind boggling how in just 20 minutes or so, I can build some stuff that would have taken weeks not too long ago, plus if there were things I didn’t understand I could clarify and it gave me really good answers.

One other thing, everybody knows, LLMs are good at travel planning. I usually prefer to plan everything myself and just get starting point from GPT, but this time we wanted to go camping spontaneously, so I asked LLM, “Lookup campgrounds within 2 hours drive from Seattle, that have plenty of first come first serve spots, access to lake, with activities including paddleboarding and biking. Create a list of 5 campgrounds with a short description.” – so it basically did all of the Googling for me based on that prompt, provided pointers to sources, etc, etc. Mindboggling.

One other thing I asked a specialized AI tool to do was to generate a 3D model to print. It did fairly good job – that’s it – giving in.

Where I think it can still be much better

There are still a few things that I want it to be better at. For example, being less of a “yes man”. Contradicting what LLMs say makes them change their mind and say “yes, you are absolutely correct, let me update the answer”. Other things are: better reasoning, understanding context even better, etc. Arguably, this would be a very tricky challenge for LLMs to be like true humans, but it appears we are definitely on that direction.

Conclusion

For me personally, LLMs and AI have now clearly crossed the threshold of being not just good enough—but genuinely useful. The time and effort it takes to engage with them are now well worth the return. Whether it’s writing code faster, mapping out life goals, or planning a camping trip, the tools have become practical enough to build into daily routines.

Having finally “given in” to their usefulness, I’m also embracing AI: having fresh curiosity and investing time to study it deliberately. It feels like the right moment to not just use the technology—but to understand it, shape how I interact with it, and grow with it.