Learning track

Working with AI

Working with AI is a skill, not a subscription. The person who gets good output does not own a better tool; they know what to ask for, where not to ask at all, and which answer not to believe.

  • 15 lessons
  • Free, no signup

The six clusters of skill this path builds

Understanding what it isWriting prompts
Privacy and businessWorking with AIsix clusters of skillDriving the tools
Detection and trustAutomating work

After the first two clusters the order stops mattering, which is why this is a comb and not a ladder. Each person starts at the cluster their own work sits in.

The path, step by step

Lessons are ordered so each stands on the previous one. You can also jump in anywhere.

  1. 1What artificial intelligence actually isArtificial intelligence is not the name of a technology, it is the name of a goal: programs that do work which until yesterday needed human judgment. What sits behind the name today is almost always a model that built its behaviour from data, rather than a rule somebody wrote in advance.Beginner
  2. 2What a large language model (LLM) isA large language model is a program that breaks text into small pieces called tokens and guesses, over and over, what the next token should be. There is no table of facts inside it; what it has are weights derived from an enormous volume of text, and those decide which continuation is more probable.Beginner
  3. 3Prompt engineering, from a prompt that worksPrompt engineering means writing four things explicitly instead of one vague request: the context, exactly what you want, what must not appear in the answer, and the shape the answer comes back in. The real skill is not the first prompt; it is reading a wrong answer and changing exactly one thing in the next round.Beginner
  4. 4Advanced prompting techniquesAdvanced prompting is a five rung ladder: zero shot, few shot, chain of thought, structured output and system prompts. A higher rung is not better, it is more; and on today reasoning models the vendors own guidance says the third rung is usually one you should not climb.Intermediate
  5. 5How to work with ChatGPTWorking with ChatGPT means opening one conversation, writing out the job you want with its details, and correcting the answer step by step, rather than asking a one-line question and judging the tool by that single reply. The real skill sits in two things: what you put inside the conversation, and what you never put there.Beginner
  6. 6How to work with ClaudeClaude is a chat-based tool built for work whose output is a long text or document edited many times over, rather than a short answer. Its free plan is listed on Anthropic own pricing page, no numeric ceiling for it is published, and Iran is on neither of that company two supported-countries lists.Beginner
  7. 7How to work with GeminiGemini is built for work whose input is a file: you attach documents, spreadsheets, images and video, and ask questions about those. It is the only one of the three big chatbots that publishes its exact ceilings as numbers, so you know where you will be stopped before you start.Beginner
  8. 8Open models: Kimi, DeepSeek and QwenAn open model means the model weights are published as downloadable files and you can run those files on hardware you control. Being open is a property of one repository, not of a company: in the Z.ai catalogue itself, GLM-5.2 sits under MIT while GLM-5.3 carries a bespoke license the vendor wrote.Intermediate
A glass brain-shaped circuit with red sparks, green checks and blue refusals floating above a desk

The order is simple. Two concept lessons first, answering what actually sits under the name and what a language model really does, then prompting and its advanced techniques, then the tools people use most and the open weight models, and last the parts that get misused hardest: agents, automation, content detection, hallucination and privacy. Do not skip the first two. Most complaints we hear about these tools come from missing those, not from a weak model.

This path is the skills layer, and the reference layer sits elsewhere: in the AI section of this site we keep what each model does, what it costs, where it lands in which ranking, and whether it has an official route from Iran, each with a date and a source. They are two places on purpose: that one changes every week, this one does not.

One thing this path does not claim: that AI does your job. In most of the tasks in these lessons the model makes the first half several times faster and the second half heavier, because output that is almost right takes longer to fix than output that is plainly wrong. Wherever that trade lands against you, the lesson says so.

Real follow-up questions

Do I need to know how to code to work with AI?

Not for any lesson in this path. What is genuinely needed is writing precisely: if you can put your request in a form a new colleague could carry out without asking you anything, you can already write a prompt. Code becomes necessary only for heavy automation, and that lesson says plainly where the job has left your hands.

Which lesson should I start with?

If you have never used these tools, start at the beginning and go in order. If you have used them and the output disappoints you, go straight to the prompting lesson and come back to the two concept lessons afterwards; that is when you find out why the prompt was failing. Anyone who only wants one specific tool can go straight to that tool lesson.

Do these tools work from Iran?

Official access is the exception, not the rule, and we neither sell accounts nor teach circumvention. What we do is this: for each tool we write down what the vendor own page says, with the date we read it, and where the vendor says nothing we write unknown. Open weight models that run on your own hardware do not carry this limit, and they get a lesson of their own.

This path is written for someone who wants to do this work themselves. If repetitive work is piling up and you would rather hand it over, see the virtual assistant page