Writing in Prothom Alo, the tech blogger and content creator Remigius Remy opens with a scene: eight o’clock on a Sunday morning, a 45-year-old woman he calls Salma holding a cup of tea, saying “AI is not for me, that’s for younger people.”
His answer is that learning AI is like learning to ride a bicycle. You only have to start once, and no particular age is required. What follows is a seven-day plan, ten minutes a day.
The plan
Sunday. Ask one easy question — something like “what should I cook today?”
Monday. Have it draft an email or an application.
Tuesday. Ask it to explain something difficult “as if to a ten-year-old”.
Wednesday. Take one piece of information the AI gave you and check it against another source.
Thursday. Rewrite an earlier question more clearly and ask again.
Friday. Look through the app’s privacy settings.
Saturday. Write down, in your own words, what you learned.
Ten minutes a day, no extra cost and no special knowledge — a smartphone and some curiosity is all it takes.
His safety note is unambiguous and correct: across the whole week, never type personal identity details, passwords, or bank and card information. Making mistakes is part of the learning; nobody is grading you.
What the plan leaves out: which app
The article never says what to open, and for a first-time user that is the step where the week stalls. Any of the major free chatbots will do — the free tier of ChatGPT, Claude, Gemini or Copilot. Install one from the official app store, not from a link in a message, and do not pay for anything in week one.
If the work is in Bangla, expect it to be noticeably weaker than in English. That is not your prompt being bad. The amount of usable Bangla training data is small, which is why AI systems still class it as a low-resource language — and every one of these tools inherits the shortage.
What “check the privacy settings” actually means
Friday is good advice stated too vaguely to act on. There are three switches worth finding, and they are in roughly the same place in every app:
Training. Whether your conversations are used to improve the model. Turn it off if the option exists.
History. Whether chats are stored, and how to delete them. Delete anything from the week you would not want read back.
Memory. Whether the app remembers things about you between conversations. Worth knowing it is on before it quotes your own details back at you.
Wednesday is the whole week
The plan treats cross-checking as one day of seven. It is the one that decides whether the other six were useful.
A chatbot produces fluent, confident text whether or not it knows the answer, and the confidence does not vary with the accuracy. A user who never checks does not learn where the tool fails — they learn to trust it uniformly, which is worse than not using it. A Scottish school hired a teacher whose entire job is teaching pupils not to believe it, which is the same lesson taught at institutional scale.
So an adjustment to Wednesday: do not pick a fact you cannot check. Ask it something you already know the answer to — a date in your own family, a local bus fare, the rules of a game you play. Catching it being wrong about something you can verify is the single most valuable ten minutes in the week.
And one addition for Saturday
Remy’s last day asks you to write what you learned. Add a second line: what you will not use it for. Travel bookings and payments are where the scam risk lives — more than 80 per cent of people now get a fake message every week — and a chatbot that confidently hands you a phone number or a payment link has given you something it cannot vouch for.
There is also a cost to getting fluent at this, which is worth naming at the start rather than discovering later: handing the thinking over has measurable effects on what you retain. Ten minutes a day to learn the tool is sensible. Ten minutes a day for the rest of your life to avoid thinking is a different habit.
