Understand how it works, use it well and think for yourself
Technology & Digital Skills4 chaptersAbout 28 min read
Kuza NdotoAI for EveryoneUnderstand how it works, use it well and think for yourselfTechnology & Digital Skills · eBook
A clear mental model of AI language models and a professional prompting framework, so AI makes you sharper, not lazier.
By Kuza Ndoto Editorial
Contents
Before you begin: why AI literacy matters now
How AI Language Models Actually Work, Without the Maths
Prompting Like a Professional: A Framework for Reliable AI Results
Your 30-day AI skills plan
Chapter 1
Before you begin: why AI literacy matters now
AI assistants have moved from novelty to everyday tool faster than almost any technology before them. Your classmates use them to revise, your future employer may expect you to use them at work, and the people who use them well are already pulling ahead. Yet most people use AI the way they would use a slot machine: type something, hope for a good result and either trust it blindly or give up. This book replaces hope with understanding and a method.
Who this book is for
This book is for students, graduates and early-career professionals who want to use AI confidently and responsibly. You do not need a technical background. You do not need to code. If you can write a clear WhatsApp message, you can learn everything here.
It solves two problems at once. The first is poor results: answers that are generic, off-target or wrong, which leave people convinced AI is useless. The second, and more serious, is poor judgement: trusting invented facts, pasting in private information, or letting the tool do your thinking until your own skills weaken. Both problems come from the same root. People use a tool they do not understand.
The big idea: understand it, direct it, check it
Using AI well comes down to three habits that build on each other.
Understand it. Know what is actually happening when an assistant answers you, so you can predict where it will be strong and where it will fail.
Direct it. Give it a clear brief, so its output fits your real situation instead of the most generic case.
Check it. Verify what matters and keep your own judgement in charge, because your name goes on the final work.
The first chapter builds understanding. The second teaches direction and checking. The final chapter turns all three into daily practice.
Key idea
AI should make you sharper, not lazier. The goal is not to hand over your thinking, but to use a fast, tireless assistant to think, draft and learn better, while you stay responsible for what is true and what is good.
A guided tour of the chapters
How AI Language Models Actually Work, Without the Maths
The first chapter gives you a clear mental model of a large language model. You will follow your question as it is split into tokens, see how the model predicts the next token again and again to build an answer, and learn how pre-training on huge amounts of text, the transformer's attention mechanism, and fine-tuning with human feedback turn a text predictor into a helpful assistant. Most importantly, you will understand why models hallucinate, stating false things with confidence. Afterwards, you will be able to predict which tasks AI handles well and which details you must always verify.
Prompting Like a Professional
The second chapter gives you a five-part framework for prompts: role and context, task, inputs, constraints and format, and examples or a quality bar. A short way to remember it is: who, what, with what, how, and how good. You will see before-and-after examples for a CV bullet, a study summary and a business email, learn how to steer a conversation through follow-up prompts, and build a routine for verifying outputs and protecting sensitive information. Afterwards, you will be able to write prompts that give reliable, usable results, and start your own prompt library.
Your 30-day AI skills plan
The closing chapter is a four-week plan that moves you from understanding to habit. You will test the model's strengths and weaknesses for yourself, practise the framework on real tasks, build a set of reusable prompts and finish with a personal set of rules for responsible use.
How to get the most from this book
Read the chapters in order. Prompting techniques make much more sense once you know why a model fills gaps with generic text and why it can invent references. The mental model from the first chapter is what turns the framework from a list of tips into something you understand.
Keep an assistant open as you read and try things. When the first chapter describes how a model can lose track in a long conversation, test it. When the second chapter shows a before-and-after prompt, run both versions on a real task of your own and compare. You will remember what you test far better than what you only read.
Keep notes in one place: a notebook or a note on your phone. Record prompts that worked, prompts that failed and why, and any errors you caught. These notes become the prompt library you build in the 30-day plan.
Watch out
Before you begin experimenting, check your university, college or employer's rules on AI use, and follow them. Never paste ID numbers, passwords, bank details, medical information or other people's personal data into a general AI tool. Use placeholders such as "Client A" when you need help with sensitive material.
One final point. AI tools change quickly, and the features of any particular product will look different in a year. The ideas in this book will not. Next-token prediction, the value of context, and the need to verify are foundations that will keep serving you whatever tool you use next.
Chapter 2
How AI Language Models Actually Work, Without the Maths
Tokens, training and attention: a clear mental model of what happens when you type a question into an AI assistant.
When you type a question into an AI assistant and a fluent answer appears a second later, it can feel like talking to a mind. It is not magic, and it is not a person. Underneath is a surprisingly simple core idea, repeated at enormous scale: predict the next piece of text. Once you understand that idea, and the training that shapes it, you will use these tools more wisely than most people around you.
Why a mental model matters
You do not need to know how an engine is built to drive a matatu, but a driver who understands brakes and fuel makes better decisions. AI assistants are the same. If you think they "know" things the way a textbook does, you will trust them in the wrong places. If you think they are just autocomplete, you will underuse them.
The truth sits in between. A large language model, or LLM, is a computer program trained on huge amounts of text to produce text. "Large" refers to its vast number of internal settings, often billions, adjusted during training. "Language model" means its job is to model how language flows. Let us walk through the pipeline, step by step, without a single equation.
Step one: text becomes tokens
Computers work with numbers, not letters. So the first thing that happens to your question is that it gets chopped into tokens. A token is a small chunk of text: sometimes a whole common word like "the", sometimes part of a longer word, sometimes a punctuation mark or a space attached to a word.
For example, "Kuza Ndoto helps students" might be split into pieces such as "K", "uza", " Nd", "oto", " helps", " students". The exact split depends on the model, because each model has its own fixed vocabulary of tokens, built from the chunks of text that appear most often. Common English words usually get one token; rarer words, names and words from less represented languages get broken into several pieces.
Each token in the vocabulary has an ID number. Your sentence therefore becomes a list of numbers, and that list is what the model actually reads.
Key idea
A model never sees your words as you see them. It sees a sequence of tokens. This explains some odd behaviour, such as models sometimes struggling to count the letters in a word: the word may be one or two tokens, not a row of individual letters.
Step two: predicting the next token
Here is the heart of it. Given a sequence of tokens, the model's only job is to estimate, for every token in its vocabulary, how likely that token is to come next. The output is a list of probabilities: perhaps "Nairobi" is very likely, "Mombasa" somewhat likely, and "banana" extremely unlikely.
The system then picks one token from that list, adds it to the end of the text, and runs the whole process again with the slightly longer text. Token by token, the answer grows. That is why you often see responses appear word by word on screen: you are watching this loop happen.
Figure 1. The generation loop. Every answer is built one token at a time by running this cycle again and again.
The system does not always pick the single most likely token. A setting often called temperature controls how adventurous the choice is. Low temperature sticks closely to the most likely options and gives predictable answers; higher temperature lets less likely tokens through, giving more varied and sometimes less accurate text. That is one reason the same question can get two different answers.
Worked example
Imagine the text so far is: "The capital city of Kenya is". Here is roughly what happens next.
The text is split into tokens and turned into ID numbers.
The model reads the whole sequence and produces a probability for every token in its vocabulary. "Nairobi" gets a very high probability, because in the training text this phrase was almost always followed by that word.
One token is chosen, almost certainly " Nairobi".
The text becomes "The capital city of Kenya is Nairobi", and the loop runs again. Now a full stop or a comma is likely next.
Notice that at no point did the model look up a fact in a database. It produced the pattern that best fits the text it had seen. Here the pattern and the fact agree. The interesting question is what happens when they do not.
Step three: pre-training on a mountain of text
How does the model learn which tokens are likely? Through pre-training. The model starts with random internal settings, so its early predictions are nonsense. It is then shown an enormous collection of text, drawn from sources such as books, articles, websites and code, and given the same task billions of times over: here is some text, guess the next token.
After each guess, the training process checks the real next token and nudges the model's settings slightly so that the correct token would have been a bit more likely. Repeat this across a vast amount of text and something remarkable happens. To get good at predicting the next token, the model has to absorb grammar, facts that appear often, styles of writing, the structure of arguments, and even patterns of reasoning, because all of these help predict what comes next.
Think of a student who has read every past paper and textbook in the library until they have an instinct for how explanations and answers usually go. That instinct is powerful, but it is an instinct about patterns, not a filing cabinet of verified facts.
Pre-training also has a cut-off. The model only learns from text gathered up to a certain date, so it does not automatically know about events after that, unless the assistant is connected to tools such as web search.
The transformer and attention, by analogy
The design used inside almost all modern language models is called the transformer, introduced in a 2017 research paper. Its key ingredient is a mechanism called attention.
Consider this sentence: "Wanjiru gave Otieno her notes because she was travelling." Who is "she"? You know instantly that it is Wanjiru. To work that out, your mind connected "she" back to the right earlier word. Attention lets a model do something similar. As it processes each token, it can look across all the other tokens in the text and decide which ones are most relevant to understanding this one.
Here is an analogy. Picture a busy group chat planning a chama meeting. When someone writes "Is it still at the same place?", you scroll up and focus on the message where the venue was mentioned, while skimming past the jokes and stickers. Attention is the model's way of scrolling up and focusing on the right messages, for every token, all at once.
A transformer stacks many layers of this, each building a richer picture of what the text means, from which words belong together up to what the overall question is asking. Because attention can connect any token to any other, transformers handle long passages far better than earlier designs, which tended to lose track of the start of a long sentence.
Try this
Ask an assistant a question that depends on something you said much earlier in the same conversation, such as a detail about your course or your budget. When it gets this right, you are seeing attention at work: it is linking your new message back to the relevant earlier text. When the conversation gets very long, models can lose track, so restating key details is a good habit.
Fine-tuning and learning from human feedback
A model fresh from pre-training is a brilliant text continuer, but not yet a helpful assistant. Ask it "How do I write a CV?" and it might continue with more questions, as if it were completing a forum page, rather than answering you.
So developers add further stages of training.
Fine-tuning on examples. The model is trained on a smaller, carefully prepared set of example conversations in which a request is followed by a good, helpful response. This teaches it the shape of being an assistant: answer the question, follow instructions, stay on topic.
Learning from human feedback. People compare different responses from the model and indicate which ones are better: more accurate, clearer, safer. A widely used approach, often called reinforcement learning from human feedback, uses these judgements to train a separate scoring system, which then guides further adjustments to the model so that it produces the kinds of responses people preferred.
Safety and policy training. Developers also train models to decline harmful requests and handle sensitive topics carefully.
The result is a model whose underlying engine is still next-token prediction, but whose habits have been shaped towards being useful in conversation.
Why models "hallucinate"
A hallucination is when a model states something false with complete confidence: an invented statistic, a made-up book title, a court case that never happened, a wrong date. Once you understand the pipeline, this stops being mysterious.
It generates what sounds right, not what is checked. The model produces the most plausible continuation. When the true answer appeared often in its training text, plausible and true usually match. When the topic is rare, local or very specific, the model may fill the gap with something that merely fits the pattern.
It has no built-in fact checker. Unless the assistant is connected to search or documents, there is no step where it verifies a claim against a source.
Fluency hides uncertainty. The writing style is equally confident whether the content is solid or invented, because confident writing is what it learned.
Its knowledge has a date. Questions about recent events, new regulations or current prices can produce outdated answers presented as current.
Watch out
Be most careful with specific details: names, numbers, dates, quotations, references, legal or medical claims, and anything about local institutions, fees or deadlines. Ask a model for a list of sources and it may produce realistic-looking titles that do not exist. Always confirm such details from an official or primary source before you rely on them.
A language model is a brilliant pattern engine, not an oracle. Use it to think, draft and explain, then verify what matters.
What this means for using AI well
Everything above translates into a handful of practical habits.
Give it context. The model predicts from the text it has. The more relevant detail you provide, such as your level, your goal and your audience, the better its predictions fit your situation.
Use it for what patterns are good at. Explaining concepts, drafting, rephrasing, summarising text you supply, brainstorming, and structuring ideas are strengths.
Supply the facts yourself when accuracy matters. Paste in the article, notes or document and ask the model to work from that. It is far more reliable when summarising text in front of it than when recalling from memory.
Verify specifics. Treat numbers, names and citations as claims to check, not facts to copy.
Iterate. If the first answer misses, tell it what was wrong and ask again. Each message changes the text it predicts from.
Protect your privacy. Do not paste ID numbers, passwords or other people's personal information into an AI tool.
This week, ask an assistant to explain a topic from one of your units, then check its explanation against your lecture notes. Notice where it is excellent and where it drifts.
Key takeaways
Your text is split into tokens, small chunks of words that the model reads as numbers.
At its core, a language model predicts the next token, adds it, and repeats until the answer is complete.
Pre-training on vast amounts of text gives the model an instinct for language and common knowledge, up to a cut-off date.
The transformer's attention mechanism lets the model link each word to the most relevant other words in the text.
Fine-tuning and human feedback turn a text continuer into a helpful assistant.
Hallucinations happen because the model produces plausible text, not checked facts, so always verify specifics that matter.
Chapter 3
Prompting Like a Professional: A Framework for Reliable AI Results
Why vague prompts get vague answers, and a five-part structure that turns AI into a dependable work partner.
Two people can use the same AI assistant on the same afternoon and walk away with completely different opinions of it. One says it is useless and generic. The other says it saves them hours every week. The difference is rarely the tool. It is the prompt: the instructions you give. This article gives you a simple five-part structure that makes your results far more reliable, plus real before-and-after examples you can adapt today.
Why vague prompts get vague answers
An AI language model produces the most likely response to the text in front of it. If your prompt is short and general, the most likely response is also general: the kind of answer that would suit almost anyone. It cannot read your mind, see your lecture notes, or know that your manager hates long emails.
Think of it as briefing a very capable new intern who started this morning. They are quick, well read and eager, but they know nothing about you, your course, your company or your standards. If you say "write something about marketing", you will get something about marketing. If you explain who it is for, what it must achieve and what good looks like, you will get something you can actually use.
Key idea
The model fills every gap you leave with the most generic option. Good prompting is simply the habit of closing the important gaps before you press enter.
The five-part prompt framework
You do not need fancy tricks or secret phrases. You need to cover five things. Not every prompt needs all five in detail, but checking each one takes seconds and prevents most disappointing answers.
Role and context. Who should the model act as, and what is the situation? For example: "You are an experienced careers adviser helping a final-year economics student in Nairobi apply for graduate trainee roles." Context includes your level, your audience and why you need this.
Task. What exactly should it do? Use a clear action verb: rewrite, summarise, compare, draft, critique, explain, list. One main task per prompt works best.
Inputs. What material should it work from? Paste the text, notes, data or draft. Mark where the input starts and ends, for instance with a line such as "NOTES START" and "NOTES END", so the model does not confuse your instructions with your material.
Constraints and format. How long? What tone? Bullet points or paragraphs? A table? British or American spelling? Anything to avoid? This is where you shape the output so it is ready to use.
Examples and quality bar. Show what good looks like, or describe it. A single example of the style you want is often more powerful than a paragraph of description. You can also say how you will judge the result: "Each bullet must include a measurable result."
A quick way to remember it: who, what, with what, how, and how good.
Before and after: three realistic tasks
Let us apply the framework to tasks you are likely to face this year. In each case, notice that the improved prompt is longer, but every extra sentence removes a guess the model would otherwise have to make.
1. A CV bullet point
Before
After
Make this CV point better: "Was treasurer of the business club."
You are a recruiter who screens graduate CVs for finance roles in East Africa. Rewrite the CV bullet below so it shows impact. Use the facts I give and do not invent any numbers. Facts: I was treasurer of my university business club for one year; I moved our records from a paper book to a shared spreadsheet; I collected member contributions and prepared a monthly report for the committee; membership grew during my term. Format: give me three alternative versions, each one line, starting with a strong past-tense verb. Quality bar: each version should show what I did and what changed as a result.
The first prompt invites the model to guess, and it may even invent impressive-sounding figures that you cannot defend in an interview. The second gives it facts, forbids invention and asks for options, so you stay in control.
2. A study summary
Before
After
Summarise photosynthesis.
You are a patient tutor helping a first-year biology student revise for an exam. Using only the lecture notes between NOTES START and NOTES END, produce a revision summary. Format: a five-sentence overview, then a bullet list of key terms with one-line definitions, then five short self-test questions with answers listed separately at the end. Keep the language simple. If something in my notes looks incorrect or incomplete, flag it rather than silently fixing it. NOTES START [paste your notes here] NOTES END
Asking it to work "only" from your notes keeps the summary aligned with what your lecturer actually taught. The self-test questions turn a passive summary into retrieval practice, one of the most effective ways to remember material.
3. A business email
Worked example
Before: "Write an email to a client saying the delivery is late."
After: "You are an operations coordinator at a small furniture business in Nairobi. Draft an email to a long-standing client, Ms Achieng, telling her that her order of six office chairs will arrive on Friday instead of Tuesday because our supplier delivered the fabric late. Goals: apologise sincerely without over-explaining, give the new date clearly, and offer free delivery as goodwill. Tone: warm, professional, confident. Length: under 120 words. Include a clear subject line. Do not promise anything beyond what I have stated."
Why it works, part by part:
Role and context: operations coordinator, small business, long-standing client.
Task: draft an email about a delay.
Inputs: the client name, the order, both dates and the reason.
Constraints and format: tone, word limit, subject line, no extra promises.
Quality bar: the three goals tell the model what success looks like.
Iterating: your first prompt is a first draft
Professionals rarely accept the first output. They treat the conversation as a collaboration and steer. The good news is that steering is easy, because the model remembers the earlier messages in the same conversation.
Useful follow-up moves include:
Be specific about what is wrong. "Too formal. Make it sound like a friendly colleague, and cut it by a third."
Ask for options. "Give me three different opening lines, from most formal to most casual."
Ask it to critique itself. "Review your answer against my quality bar and list anything it misses, then revise."
Ask it to ask you. "Before you write, ask me up to five questions about anything you need to know." This is excellent for bigger tasks such as a cover letter or a project plan.
Start fresh when it drifts. In a very long conversation, the model can lose track of early instructions. Open a new chat and paste a clean, complete prompt.
Try this
Keep a small personal prompt library in your notes app. Each time a prompt gives you an excellent result, save it with blanks you can refill, for example: "You are a [role]. Using only the text between INPUT START and INPUT END, [task]. Format: [format]. Quality bar: [standard]." Within a month you will have a toolkit for your most common tasks.
Verifying outputs before you use them
A well-structured prompt improves results enormously, but it does not make the model infallible. Language models can state false information confidently, especially specific details such as names, dates, statistics and references. Your name goes on the final work, so the responsibility stays with you.
A simple verification routine:
Check every fact you did not supply. If the output contains a number, a date, a law, a quote or a source that was not in your input, confirm it from an official or primary source, or delete it.
Check it against your brief. Did it follow the format, length and tone? Did it answer the actual question?
Read it aloud. AI text can sound polished but empty. Cut filler phrases and make sure it sounds like you.
Follow your institution's rules. Universities, colleges and employers have policies on AI use. Know them, and be honest about how you used the tool when asked.
Watch out
Never paste sensitive information into a general AI tool: ID or passport numbers, passwords, PINs, bank details, medical records, or a colleague's or client's personal data. The same goes for your employer's confidential documents. Many services may store conversations, and some may use them to improve their systems depending on your settings. When you need help with sensitive material, replace real names and numbers with placeholders such as "Client A" and "Amount X", or check whether your organisation provides an approved tool.
Common mistakes to avoid
Asking several unrelated things at once. Split big jobs into steps: outline first, then draft, then polish.
Hiding the real goal. "Write a paragraph about teamwork" is weaker than "Write a paragraph for my internship application that shows I can work in a team, using this example from my group project."
Letting it invent your achievements. Always tell it to use only the facts you give.
Copying without editing. The output is raw material. Your judgement is what makes it good.
Giving up after one try. A poor first answer is information. Adjust the prompt and go again.
What to do this week
Pick one real task you are facing: a CV update, a set of revision notes, or an email you have been putting off. Write a one-line prompt and save the result. Then rewrite the prompt using all five parts of the framework and compare the two outputs side by side. Notice which part made the biggest difference. Save the better prompt as the first entry in your library.
Do this three or four times and the framework becomes a habit. You will stop thinking of AI as a slot machine and start treating it like what it can be: a fast, tireless assistant that is only as good as the brief you give it.
Key takeaways
Vague prompts get generic answers because the model fills every gap with the most likely, most general option.
Use the five parts: role and context, task, inputs, constraints and format, and examples or a quality bar.
Give the model your facts and tell it not to invent anything, especially on CVs and applications.
Treat the first output as a draft, and steer with specific follow-up instructions.
Verify every fact you did not supply, and keep sensitive personal or company data out of AI tools.
Chapter 4
Your 30-day AI skills plan
Understanding AI is the start. Using it well every day is the skill. This 30-day plan turns the mental model and the prompting framework from this book into habits you will keep. Plan for 20 to 40 minutes a day, using real tasks from your studies or work, and follow your institution's or employer's rules on AI use throughout.
Week 1: Test the machine for yourself
This week applies the chapter on how language models work. You will see tokens, prediction and hallucination with your own eyes, so your trust is based on evidence.
Day 1: Set up your notes. Create an "AI log" in your notebook or phone with three headings: what worked, what failed, and errors caught.
Day 2: Explain and compare. Ask an assistant to explain a topic from one of your units, then check it against your lecture notes. Record where it was excellent and where it drifted.
Day 3: Watch the variation. Ask the same open question twice in fresh chats. The answers differ because the model chooses among likely tokens.
Day 4: Probe for hallucination. Ask for sources or specific facts on a narrow, local topic you know well. Check every name, date and title. Log what was invented.
Day 5: Test attention. In one conversation, mention a detail early, then ask a question much later that depends on it. Notice when it links back correctly and when it loses track.
Day 6: Supply the facts. Paste in a short article or your notes and ask for a summary using only that text. Compare its accuracy with Day 4, when it worked from memory.
Day 7: Review. Write three rules of thumb about where you will and will not trust AI, based on this week's results.
Week 2: Practise the five-part framework
This week applies the prompting chapter. You will practise each part of the framework until it becomes automatic.
Day 8: Before and after. Pick a real task. Write a one-line prompt and save the result. Rewrite it using all five parts: role and context, task, inputs, constraints and format, and quality bar. Compare the outputs side by side.
Day 9: Focus on inputs. Give the model your own material, clearly marked with start and end lines, such as NOTES START and NOTES END. Ask it to work only from that material.
Day 10: Focus on format. Take one piece of content and ask for it three ways: a table, five bullets and a 100-word paragraph. Notice how much control the constraints give you.
Day 11: Focus on the quality bar. Give the model one example of the style you want, or a clear standard such as "each bullet must show a result". Compare with a version that has no quality bar.
Day 12: Protect your facts. Rewrite a CV bullet using the chapter's approach: supply the real facts and tell the model not to invent any numbers. Ask for three alternative versions.
Days 13 and 14: Turn notes into retrieval practice. Ask for a revision summary from your own notes, with self-test questions and answers listed separately. Answer the questions without looking.
Week 3: Steer, iterate and build your library
This week focuses on treating the first output as a draft, as the prompting chapter recommends.
Day 15: Be specific about what is wrong. Take a weak output and give precise feedback, such as "too formal, cut it by a third". Log how many rounds it took to reach something usable.
Day 16: Ask it to ask you. For a bigger task, such as a cover letter or project plan, start with "Before you write, ask me up to five questions". Compare the result with your usual approach.
Day 17: Ask it to critique itself. Have the model review its own answer against your quality bar, list what it misses, then revise.
Day 18: Start fresh when it drifts. In a long conversation that has wandered, open a new chat with one clean, complete prompt. Note the difference.
Day 19: Build template one. Save your best prompt so far with blanks you can refill, such as [role], [task], [format] and [standard].
Days 20 and 21: Build templates two to five. Create templates for your most common tasks: revision notes, emails, CV bullets, summaries of readings, brainstorming.
Week 4: Verify, protect and think for yourself
The final week strengthens judgement, using the verification routine from the prompting chapter and the hallucination lessons from the first chapter.
Day 22: Run the full verification routine. On one output, check every fact you did not supply, check it against your brief, read it aloud and cut filler.
Day 23: Practise placeholders. Take a task involving personal or confidential details and redo it with placeholders such as "Client A" and "Amount X" instead of real data.
Day 24: Think first, then compare. Before asking the AI, write your own outline or answer to a problem. Then ask the model and compare. Note what you saw that it missed, and the reverse.
Day 25: Use it as a tutor, not a ghostwriter. Ask it to quiz you or to explain a hard concept more simply, rather than producing the work for you.
Day 26: Write your personal AI rules. Five lines covering what you use AI for, what you never use it for, what you always verify, what you never paste in, and how you disclose its use when asked.
Day 27: Teach someone. Explain next-token prediction and the five-part framework to a friend in five minutes. Teaching reveals gaps in your own understanding.
Days 28 to 30: Real-world project. Use AI on one meaningful task from start to finish, such as a job application or a revision pack, applying the whole method. Review your AI log and note your three biggest improvements since Day 1.
Your day-30 checklist
I can explain, in plain language, how a language model produces an answer one token at a time.
I have caught and logged at least three AI errors or invented details.
I write prompts using all five parts of the framework without needing to look them up.
I have at least five saved prompt templates for tasks I do often.
I routinely verify every fact in an output that I did not supply.
I use placeholders whenever a task involves sensitive information.
I have written personal AI rules and know my institution's or employer's policy.
Watch out
The most common way people fall off this plan is not by stopping AI use, but by sliding back into one-line prompts and copy-paste habits once the novelty wears off, usually during a busy week. Under pressure, the framework feels slow, so it gets skipped, and errors creep back in. Protect yourself with your template library: a saved prompt takes seconds to fill in. And keep Day 24's habit of thinking first, so the tool supports your thinking rather than replacing it.
Key takeaways
Test the model yourself, so your trust in it rests on evidence rather than impressions.
Supply your own material when accuracy matters; models are far more reliable working from text in front of them.
Use the five parts every time: role and context, task, inputs, constraints and format, and quality bar.
Treat every first output as a draft, and steer with specific follow-up instructions.
Build a prompt library so good habits survive busy weeks.
Verify what you did not supply, protect sensitive data, and think first so AI sharpens your mind instead of replacing it.
Keep learning
Turn what you've read into skills with a structured course or a free tutorial series.