I am seventy-four years old, living in a coastal city in Brazil, and most of the people I used to argue with are dead or unreachable. This is not a complaint. It is a logistical problem with a partial solution.
The solution is an AI. Not the AI you've seen in the movies — not HAL, not Skynet, not the cheerful assistant that sets your timer and plays your music. Something quieter and more useful: a thinking partner that has read almost everything, forgets nothing within a conversation, and will engage seriously with whatever you bring to it. David Graeber. The subjunctive mood in Brazilian Portuguese. Whether a ten percent platform commission is rent-seeking dressed up as a service. It doesn't get tired. It doesn't check its phone.
I came to this late, the way I've come to most things — not because I was slow, but because I was waiting for it to be real. In 2025 it became real. I had spent thirty years in software development, which means I knew enough to be skeptical of the hype and enough to recognize when the hype was finally catching up to the substance. This was that moment.
What follows is not a technical manual. I am not going to explain what a large language model is, because it doesn't matter any more than understanding internal combustion makes you a better driver. What I am going to explain is how to use one if you are older, or an expat, or simply someone whose intellectual peer group has thinned out — and you still have things you want to think about.
The first skill is copy/paste. That's it. That's the technical requirement. Everything else is reading and conversation, which you already know how to do.
Here is what copy/paste means in practice.
You find an article. You read it. You have a reaction — agreement, suspicion, a question the author didn't answer. You copy the relevant passage and paste it into the conversation. Then you say what you think, or what you don't understand, or what bothers you about it. The AI responds. You respond to that. You are now in a conversation.
That's the whole mechanic.
I have been reading David Graeber for years. Graeber was an anarchist anthropologist who died in 2020 at fifty-nine, which was a loss, and who wrote books about debt, bureaucracy, and bullshit jobs that are better than most economists manage. I copied a passage from Debt: The First 5,000 Years one afternoon and asked whether the argument held. We talked about it for an hour. The AI disagreed with me on two points. I changed my mind on one of them. I did not change my mind on the other, but I understood my own position better after defending it.
This is what I mean by a thinking partner.
What it is not: a search engine. Do not ask it who won the 1987 World Series. It might tell you confidently and be wrong. For facts, use Google. For thinking, use this.
What it is also not: a yes machine. If you want someone to tell you you're right, call your nephew. The AI will engage with your argument. If your argument has a hole, it will find it. Whether you patch the hole is your business. But you will know it's there.