← The study library

Field-guide course · New to AI

Starting from zero

Your first assistant, your first conversations, and the two limits to learn on day one.

6 lessons · For: You have heard about ChatGPT, maybe typed a question into it once. Nothing here assumes prior experience. · Product details last reviewed July 2026

Every proficient AI operator started exactly where you are: staring at an empty chat box, unsure what these tools are actually for. This course gets you from there to a daily working habit in six short lessons — which product to start with, what to type, what to trust, and when paying makes sense.

One promise before you begin: the skill you build here transfers. ChatGPT, Claude, Gemini, and the rest differ in features and character, but the craft of working with them is one craft. Nothing you learn on your first assistant is wasted if you later switch.

  1. Lesson 1

    What you are actually talking to

    An AI assistant is a language model: a system trained on an enormous amount of text until it became remarkably good at producing the next words — an answer, a draft, an explanation — that fit what came before. That one mechanism turns out to cover a startling range of work: drafting and rewriting, summarizing long documents, explaining anything at any level of depth, brainstorming, translating, turning messy notes into clean structure.

    Two limits matter from day one. First, the model does not know your situation — your project, your inbox, your company — unless you tell it in the conversation. Second, it can be confidently wrong: when it lacks the facts, it produces something plausible instead of going silent. Neither limit is a reason to stay away. Both are reasons to learn the craft rather than just the tool.

    Try it this week

    Ask an assistant three questions: one about general knowledge, one about something you know deeply, and one about your own work. Notice where it shines, where it fumbles details, and where it simply cannot know — that map is the whole course in miniature.

  2. Lesson 2

    Meet the assistants

    Four products cover almost everyone's starting point, and every one of them has a genuinely useful free tier. The right move is not to research for a week — it is to pick one, use it daily for two weeks, and let experience form your opinion. You are choosing a first instrument, not a spouse.

    A few others are worth knowing by name even if you do not start there: Grok (from xAI, built into X, strongest on real-time chatter and current events, with a deliberately unfiltered personality), Meta AI (free and already inside WhatsApp and Instagram), and Perplexity (a search engine reimagined as an assistant — every answer arrives with citations).

    In the toolbox

    • ChatGPT

      OpenAI · Free tier; Plus ~$20/mo

      The household name and the broadest toolbox: strong general chat plus voice conversation, image generation, document upload, and research modes. The safest default if no other factor pulls you elsewhere.

    • Claude

      Anthropic · Free tier; Pro ~$20/mo

      Widely preferred for the quality of its writing and its care with long documents and nuanced instructions. If your work is mostly words — drafting, editing, thinking through hard questions — start here.

    • Gemini

      Google · Free tier; AI Pro ~$20/mo

      Google's assistant, woven through Gmail, Docs, Drive, and Android, with a generous free tier. If your life already runs on Google, Gemini meets you where your email and files already live.

    • Microsoft Copilot

      Microsoft · Free tier; bundled into many work Microsoft 365 plans

      The assistant inside Word, Excel, Outlook, and Teams. At many workplaces it is the one AI tool already approved and paid for — which can make it the right first choice by default.

    Try it this week

    Create a free account on one assistant today. Put the app on your phone's home screen and the website in your bookmarks bar — friction is the main thing that kills a new habit.

  3. Lesson 3

    Your first ten conversations

    The empty chat box intimidates because it promises everything and suggests nothing. So start with tasks, not questions: rewrite an awkward email before you send it; paste a dense document and ask for a plain-language summary; plan a trip with real constraints (dates, budget, a picky eater); turn what is in your fridge into a week of dinners; ask what questions to bring to a medical appointment or a car mechanic; have a concept you half-understand explained, then explained again more simply.

    The deeper lesson inside all ten: this is a conversation, not a search box. When the first answer is 70% right, do not start over — reply. "Shorter." "Warmer." "The second paragraph misses the point — the delay was the vendor's fault, not ours." The follow-up message is where the real value lives, and it is the single habit that most separates people who get something from these tools from people who bounced off them in 2023.

    Try it this week

    Run five of those starter tasks this week, and send at least one follow-up message in each conversation before you accept the result. The follow-up is the skill.

  4. Lesson 4

    When it is confidently wrong

    Sooner or later — usually sooner — your assistant will state something false with perfect fluency. The industry calls these hallucinations. They cluster predictably: precise numbers, names and dates, citations to articles and studies, and anything that happened recently. The tone never wavers, which is exactly the problem: confidence is not evidence.

    Two habits neutralize most of the risk. First, anything you would repeat to another person or act on, verify against a real source — most assistants can search the web on request, and asking for sources (then actually opening one) takes a minute. Second, know that assistants agree too easily: they are trained to be helpful, which shades into telling you what you seem to want to hear. When a decision matters, ask "what is the strongest case against this?" and see if the answer changes.

    Try it this week

    Ask your assistant about a topic you know deeply — your profession, your hometown, your hobby. Hunt for the subtle error or the too-smooth agreement. Finding one, personally, teaches calibration better than any warning label.

  5. Lesson 5

    Free and paid, plainly

    The free tiers are real products, not demos — for a few conversations a day they may be all you ever need. Their limits are usage caps that appear at busy moments and quieter access to each company's strongest models. The ~$20-per-month tier (ChatGPT Plus, Claude Pro, Google AI Pro, and equivalents) is the workhorse upgrade: the best models as your default, far higher caps, and the premium features — deep research modes, more generous file handling, early access to new capabilities.

    Above that sit $100–250 power tiers (ChatGPT Pro, Claude Max, Google AI Ultra) built for people whose entire working day runs through these tools. The honest advice for a beginner: stay free until you hit the walls, and let the walls make the decision. If you bump into caps twice in one week, or keep wishing for better answers on work that matters, the $20 pays for itself in the first saved hour.

    Try it this week

    For one week, jot a note each time you hit a usage cap or catch yourself wishing the answer were smarter. Decide about upgrading with that data, not with a subscription page's persuasion.

  6. Lesson 6

    Make it a default, not an event

    The gap between people who "tried AI" and people it genuinely helps is not talent — it is that the second group made asking a reflex. The reflex has a trigger: any task that is made of words and would take more than ten minutes gets thirty seconds of "could the assistant do the first draft?" before you start manually. The draft will not be finished work — finishing is your job — but starting from a draft instead of a blank page is where the hours come back.

    Two accelerants: put voice to work — the mobile apps' voice modes (ChatGPT's and Gemini's are excellent) turn commutes and dish-washing into thinking time — and keep a running note of wins and failures. Ten entries in, you will have something better than any listicle: a map of what AI is actually good for in your life.

    Try it this week

    For one week, before starting any writing-shaped task, spend thirty seconds asking your assistant first. Log each attempt as time-saved or time-wasted. Most people's log settles around 70/30 — and the 30 teaches as much as the 70.

Where next

You now have a working assistant and working habits. The next course, The everyday assistant, takes each part of your week — writing, research, documents, meetings, images — and shows you the specific tools built for it.

The products above change; the judgment the exam scores does not. The study guides teach that durable layer, and the free check scores you on it in about fifteen minutes.