The best way to learn is by doing

Ever since the AI boom started, I've watched a huge number of people create content about it.

First came the explainer videos, the ones that helped you get your bearings in an industry that ships something new every week. Then came the social media content telling you everything you could do with it, so you'd get the general idea and not waste time on tools or solutions that wouldn't add anything to your work or your life.

Now it's the moment of guides and courses. I made one myself, an introduction to AI guide, and anyone who wants it can grab it from the resources section. Courses also promise practicality and usefulness. But you can't help asking: in a year, are you going to need another one to stay current?

We got used to everything changing

I think we're going through a phase of big shifts. Younger people are almost more used to things changing constantly at a technological, social, political and economic level than to things staying the same.

That gives us a real capacity to adapt to new environments and to look at change from a positive angle. It also makes us more impatient, hungrier for fast dopamine, and less patient about the long term and about the fact that solid, good things take a long while to build.

We didn't build this

The other day I was listening to a podcast in Portuguese (trying to learn it, haha) about Gen Z. Among other things, they were talking about the global housing crisis facing young people, both in terms of what we can afford and what's even available. And about how we get criticized quite a bit for a lack of commitment or for having a short-term mindset, something that social media amplifies.

What I liked was what they said next: we're the generation that spends the most hours a day on a smartphone, but we didn't create social media. We didn't create smartphones either, and we didn't think that constant comparison, conscious and unconscious, was going to make us feel bad about our own lives sometimes.

And just this week, on Wednesday August 26, Meta reached a settlement of around 17 billion dollars with 47 US states, plus Washington D.C. and three territories, ending a trial that had just started in Oakland. The original lawsuit was filed by 29 states back in 2023, and the core accusation was exactly that: that Facebook and Instagram were designed to hook teenagers and that the company downplayed the risks.

It's worth being precise here, because it isn't a fine. It's an out-of-court settlement, and Meta admits no wrongdoing. Of that money, roughly 12.1 billion is firm and gets paid out over ten years; the other 5 billion is only owed if Snap, TikTok and YouTube agree to comparable terms. What Meta does accept are concrete measures: a default two-hour daily limit on teen accounts, a nighttime block only parents can lift, and hiding "like" counts.

I don't think this is only about Instagram. In The Social Dilemma, the Netflix documentary that came out in 2020, they laid out pretty clearly how the teams behind these platforms were incentivized to design apps that kept users around as long as possible, from the vibrant color of a notification to manufacturing FOMO. When a product is free, it's because you're the product.

From business intelligence to something that feels off

This keeps happening to me. I talk about something, and it used to be ads on Instagram. Fine, the excuse was reasonable: more personalized ads, better user experience. But now I talk about something and content about exactly that shows up on TikTok. Not ads, content. I download an app and content about that same app shows up.

I want to be honest here about what I know and what I don't. There's no public evidence that Instagram or TikTok record what you say through your microphone, and the independent research that went looking for it didn't find it. The one time a company put it in writing was Cox Media Group, a US media group that in 2024 was selling advertisers a product called Active Listening, bragging about exactly that. When 404 Media published the materials, the company pulled them and denied listening to conversations. What is documented is that with cross-app tracking, location, your contact list and what the people around you do, you can infer so much that the result feels an awful lot like being listened to.

And that's where something shifts for me. Reading Small Data, by Martin Lindstrom, I kept turning this over. His argument is that the small clues you observe in a customer's home are worth more than mountains of aggregated data, and it made me think about the opposite case, the one that always comes to mind: Target, the American retail chain. Their data team built a score, based on about 25 products, that estimated the probability a shopper was pregnant. And it worked. The most famous anecdote from that story, the angry father who finds out from the coupons arriving at his house, is pretty questionable these days, but the model itself was real.

To me, that's still business intelligence: you're using your own purchase data to draw strategic conclusions about your business. When the plan becomes inferring what you talk about from everything surrounding your phone, it stops being business intelligence and starts to look like surveillance. That's where freedoms get squeezed.

I went off on a tangent there, but I didn't want to skip this, and I'd like to hear what you think if it interests you.

What I think is going to matter most now

I think daring to try things and actually build has never been as accessible or as cheap as it is today. With a computer, a 20 dollar subscription (or more, for whoever wants to pay it) and a curious mind that wants to experiment, there are very few structured courses worth more than that.

What used to look unusual, being self-taught, seems essential to me today for anyone who wants to do something meaningful personally or professionally. In fact, if I had to start learning AI today, that's where I'd start.

A very clear example. I've kept a gym routine going for five straight months, the longest stretch of my life. AI made the path simpler for me on exercises, on nutrition and on learning. I'm still nowhere near a fitness expert. But what I learned about how to train to reduce the back pain that comes from sitting for hours, how to build muscle and how to lose weight, something that might have taken me six months before, I had fully adapted to me in three at most thanks to AI.

This applies to anything. I've run several experiments by now: finding podcasts about what interests me, YouTube channels, courses, books. Also getting into fields that always caught my attention but that I never knew how to start with. I think anyone with a minimum of proactivity can feel better today doing very little.

The judgment is still yours

I don't let an algorithm, an LLM or other people decide for me. Or at least that's what I believe. I try to keep my own judgment, to think critically and to question what I'm learning, because we went from learning alone to having a very intelligent assistant that gets things wrong sometimes and that wants to please us by resolving our doubts fast. I wrote up how I do it step by step in the 4-step method I use to learn any skill with AI.

To be clear, I'm not saying use AI for everything, because creating new dependencies or possible addictions isn't the plan either. After everything above, that would be pretty incoherent of me. But do use it for certain things, and do give yourself permission to experiment in this huge, wide world of opportunities that life is.

And if you're wondering where to start, the real risk isn't picking the wrong tool, it's not experimenting at all.

What do you all think about this? I'm genuinely curious, especially about the social media part.

Thanks for reading. Un abrazo.