Coming Soon
This feature isn't live yet. Check back soon.
Buy a hard copy
$101 each, shipping included.
Edition 2027-1About this Book
You're about to enter a job market where everyone assumes you already know how to use AI.
Almost nobody was actually taught.
Why this book exists
Most people learn AI by poking at it. They type a question, get a decent answer, and stop there. They never find out what it's actually good at, where it quietly fails, or how to tell the difference. That's not a skill gap you can afford for long.
This book teaches the ideas underneath the tools. Not which app to download, those change every few months, but how these systems work, what they can and can't do, and how to use them without embarrassing yourself.
Every lesson ends with something to actually try. Reading about AI teaches you almost nothing. Using it deliberately, and checking what happened, teaches you a lot.
What this book is not
- A list of tools to install. Tools change, principles don't.
- A shortcut around doing your own work.
- A promise that AI is always the right answer. Sometimes it's the wrong tool entirely.
- A one-time read. It's built to be a reference you come back to.
About the smiley face
Some lessons carry a in the contents. It means a curious twelve-year-old could follow the explanation. It does not mean the idea is a small one.
Those are usually the pages we rewrote the most. Anyone can make a hard idea sound complicated; making it land in plain language is the harder job, and it is the one worth doing. If a page carries the smiley, it is the clearest version we could write, not the shallowest.
Meet Bit, our helper
That's Bit. He turns up in the case study inside most lessons, freshly hired by a business with real work to do. A clinic puts him on the phones, a print shop hands him the design work, a newsroom lets him draft. Sometimes he nails it. Sometimes he is confidently wrong, and that is the lesson.
How to get through it
You don't have to read this in order. Skim the contents, find the thing that sounds like a problem you actually have, and start there. Each lesson stands on its own.
Contents
Start wherever the problem sounds like yours. Nothing here has to be read in order.
Read this first
Core skills
Bonus
Foundations
Prompting & Communication
Thinking Critically & Verification
Writing & Editing
Research & Learning
Images, Audio & Video
Data & Analysis
Coding & Building
Study & Productivity
Career & Job Search
Business & Entrepreneurship
Ethics, Safety & Responsible Use
Creative & Personal Projects
Advanced & Agentic Workflows
Who's Teaching This
Who I am
I'm Mr. Z. I build things with AI for a living and I teach people how to do the same. I've watched a lot of smart people either write AI off entirely or trust it way too much, and both mistakes cost them.
This book is the version of AI education I wish existed when people first started asking me how to use this stuff.
What I believe about learning this
- You learn by doing, not by reading about doing
- Understanding why something fails beats memorizing what to type
- Skepticism is a skill, not an attitude
- The goal is to think better, not to type less
About VibeCraft
VibeCraft is where this comes from. The idea behind it is simple enough to put on the cover: build with purpose. Use the tool because it makes the work better, not because it's the tool everyone's talking about.
A note on tools
You won't find brand names in most of these lessons. That's on purpose. The chat assistant you use this semester may not be the one you use next year, but the way you ask it questions, check its answers, and decide whether to trust it will transfer to whatever comes next.
What People Mean When They Say “AI”
The word covers about six different things, and mixing them up is the fastest way to get confused.
The short version
When most people say AI right now they mean a large language model: a system trained on huge amounts of text that predicts what words come next. That's what a chat assistant is.
It also covers image generators, speech recognition, recommendation feeds, and self-driving software. These work and fail differently.
The systems have converged: the same ideas you learn here about prompting, context, and probability now explain recommendation feeds and agents too, not just chat.
None of these understand what they are saying the way you do. They are very good at patterns, which is enough to be useful and enough to be confidently wrong.
Things all of these share
- They learn patterns from data instead of following written rules
- They produce likely answers, not guaranteed correct ones
- They reflect their training data, blind spots included
- They get better at narrow tasks, not at understanding
The main kinds you'll run into
Chat assistants
Answer questions, explain, draft, and rewrite. The default thing most people mean by AI.
Reasoning models
Work through a problem in steps before answering. Slower, far better at maths and logic.
Coding assistants
Write, explain, and fix code. They sit in the editor and see your whole project.
Agents
Don't just answer, they act. They browse, run code, and chain steps toward a goal.
Image, video, and audio generators
Make pictures, clips, music, or a voice from a description. Very different failure modes.
Everything running quietly
Search ranking, recommendation feeds, fraud detection, autocorrect. Used constantly, rarely called AI.
How to Use This Book
Find the sentence below that sounds most like you, and start there.
Where to start
“I've barely used AI at all.”
Start with the Foundations lessons. They're short and they make everything else make sense.
“I use it constantly but get mediocre results.”
Go straight to Prompting. That's almost always the gap.
“I got burned by something it made up.”
Thinking Critically. Start with checking accuracy.
“I'm worried about getting in trouble for using it.”
Read Using AI Without Cheating on page 10 before anything else.
“I want to build something with it.”
Work through Foundations and Prompting first, then the Workflow lessons.
“My professor said to use it and I don't know how.”
How to Use This Book, then the Sample Lesson on page 12 to see the format.
How each lesson is built
Read the hook
One sentence telling you what problem the lesson solves. If it doesn't sound like a problem you have, skip it.
Learn the idea
A short explanation plus where you'd actually use it.
Pick your level
Beginner, Intermediate, or Advanced, with roughly how much practice each takes.
Do the assignment
Every lesson ends with something to try. This is the part that actually teaches you.
The assignments are the point
You can read this entire book in an afternoon and learn almost nothing. The lessons are short specifically so you spend your time doing the exercises instead of reading about them.
How to Tell Good Output From Bad
Fluent writing feels like correct writing. Your brain will not flag the difference for you, so you have to do it on purpose.
The core problem
These systems are optimized to produce text that reads well. They are not optimized to produce text that's true. Most of the time those overlap. When they don't, nothing in the output looks different.
So you can't judge by feel. You need actual criteria, applied deliberately, especially on anything you'll submit or act on.
Signals something is off
- Specific numbers or dates with no source
- Quotes attributed to real people you can't find
- Suspiciously tidy lists where reality is messy
- Confident answers about very recent events
- Anything that happens to confirm exactly what you hoped
A quick quality check
Is it specific enough to be wrong?
Vague claims are safe and useless. Specific claims are useful and checkable. Prefer output you can actually verify.
Would an expert agree?
If you don't know, that's your answer about whether to use it unverified.
Does it answer what you asked?
Models often answer a nearby, easier question. Re-read your prompt and check.
The one habit worth building
Before using anything, ask yourself out loud: what in here would be embarrassing if it turned out to be false? Then check those parts specifically. It takes two minutes and it's the difference between using AI well and getting caught out by it.
Picking a Tool and Paying for It
If you're learning, free is genuinely fine. If you're building things, one good paid tool beats four mediocre ones.
Free is fine, until it isn't
For learning concepts, asking questions, checking your understanding, and most schoolwork, free tiers are completely adequate. You can do most of this book without spending anything, and you should start there.
Where free starts to hurt is building. Ask a weak model for a working app and you'll get something broken, then conclude AI is overhyped. That's the wrong lesson, learned from the wrong tool.
Signs you've outgrown free
- Your builds keep coming back broken or half-finished
- You hit usage limits inside a single work session
- You need long documents understood all at once
- You're comparing four tools instead of getting good at one
If you're going to pay, pay for one
One tool, not four
Four subscriptions because each is slightly better at one thing is the most common and most expensive mistake.
Rotate monthly until it fits
These are month to month. Spend a month with one, then switch. After three months you'll know which suits how you work, having paid for one at a time.
Check your school first
Many universities give students paid access free. Ask before you buy something you already have.
What paying does not buy
A paid plan does not make the output true. Paid models still invent sources, still have a knowledge cutoff, and still need checking. Money solves capability and capacity problems. It does not solve judgement problems, and judgement is what this book is actually about.
Questions to Ask Before You Trust It
A short list you can run in under two minutes, before anything gets submitted or acted on.
Ask about the claim
Where would this information have come from?
If the answer is “nowhere it could plausibly know,” stop. Recent events and private information are common failure zones.
Is this checkable?
Names, numbers, dates, and quotes can be verified. Vague generalities can't, which is its own kind of warning.
Did I ask a leading question?
If you asked it to argue a position, it will, whether or not the position holds up.
Would I bet money on this?
A surprisingly effective gut check. Most of the time the honest answer is no.
Ask about yourself
Do I know enough to spot an error?
If you can't evaluate the answer, you can't use it unverified. That's the whole rule.
Am I using this to think, or to avoid thinking?
One of these makes you better at your work. The other makes you dependent.
What happens if this is wrong?
Low stakes, move on. Graded, published, or acted on, verify it.
If you only remember one
“Would I bet money on this?” catches most of what the other questions catch, and you'll actually remember to ask it.
Using AI Without Cheating
This is the lesson most likely to save you from a very bad meeting with your department.
The honest framing
There is no universal rule. Policies differ by school, by department, and often by individual professor for the same course. Anyone who tells you a blanket answer is guessing.
What doesn't change: submitting work you didn't do and can't explain is the thing that gets people in trouble. Everything else is negotiable and usually written down somewhere you haven't read yet.
Almost always fine
- Explaining a concept you're stuck on
- Generating practice questions to test yourself
- Getting feedback on a draft you wrote
- Debugging code you're trying to understand
- Brainstorming before you commit to a direction
Risky, know the policy first
Drafting
Having AI produce a first draft you then heavily revise. Some courses allow it, some call it plagiarism. Ask.
Editing and rewriting
Usually fine for grammar and clarity. Less fine when it substantially rewrites your argument.
Submitting output as yours
Almost never acceptable, and easily the most common way people get caught.
The test that actually works
Could you explain every sentence you're submitting, in your own words, if asked on the spot? If yes, you're almost certainly fine regardless of how you got there. If no, you have a problem, and it's the same problem whether AI wrote it or a friend did.
Stacking Tools Together
Most of the impressive things people do with AI aren't one clever prompt. They're four ordinary steps in a row.
The idea
A single tool doing a single thing is where everyone starts. The jump in usefulness comes from chaining: use one tool to produce something, another to check it, a third to reshape it into what you actually needed.
You don't need special software for this. Copy and paste between tools is a workflow. It just has to be deliberate rather than accidental.
Common combinations
- Brainstorm in a chat assistant, then verify with actual search
- Draft with AI, edit yourself, then have AI check your edit
- Turn lecture notes into practice questions, then grade yourself
- Generate an outline, write it yourself, ask for a critique
A worked example
Get the raw material
Ask for ten possible angles on your essay topic. Most will be generic. Two might be interesting.
Filter with judgment
You pick the angle. This is the step people skip, and it's the step that makes the work yours.
Build it out
Ask for counterarguments to your chosen angle. Now you're using it to stress-test rather than to generate.
Verify separately
Check any factual claims against real sources before they go anywhere near your draft.
Where stacking goes wrong
Feeding AI output back into AI repeatedly, with no human judgment in between, produces confident nonsense that drifts further from reality at every step. The point of stacking is that you're in the chain, not that you've removed yourself from it.
Sample Lesson
One sentence naming the problem this lesson solves, in the words a student would actually use.
The idea
A short, plain explanation of the idea. Two paragraphs at most, no jargon, written like a person talking rather than a textbook.
The second paragraph usually says how you'd apply it, and points to the next page.
Where you'd use this
- The situation where you'd reach for this
- The kind of assignment it helps with
- A moment where not knowing it would cost you
- Anything that tells you this lesson is or isn't for you
Skill levels
Work up from wherever you areBeginner
Know it
- The minimum worth understanding
- One idea you could explain to a friend
- Enough to stop making the obvious mistake
Intermediate
Use it
- What most people should aim for
- Applying it to your own work
- Recognizing when it isn't working
- Adjusting instead of starting over
Advanced
Teach it
- Explaining it to someone else
- Knowing the edge cases
- Predicting where it will fail
Build this · Case Sample: The Pattern
Every case describes a real-feeling business and a process where AI is quietly involved. Bit is usually somewhere nearby, watching it happen.
Step 1, state the problem: state the problem in one sentence, in your own words, before you build anything.
Step 2, pick your goal: every case offers the same two, in that business's own words. More money in, or less money out? The story sets up both. You choose.
Step 3, build a fix: whatever moves the goal you picked. A product, a marketing piece, a workflow, a policy; the medium is entirely up to you.
Sample Lesson
Lesson detail
How to do it
First step
What to do first. Steps are short and in order.
Second step
Usually the part where you actually try something.
Third step
Often where you check or compare your result.
Fourth step
Ends with something you keep or can point at.
Key takeaways
First takeaway
The thing everyone should leave with, colored to the Beginner level.
Second takeaway
What you get once you've actually practiced it.
Third takeaway
The insight that only shows up once you can teach it.
How to know you've got it
A question you can answer yes or no
Written so you can honestly self-assess without a quiz.
A thing you can now do
Concrete and observable, not "feel confident about."
A mistake you stopped making
Often the clearest evidence something landed.
A cost or a gain you can name
What the skill saves, or what it brings in.
Common mistakes
The obvious first mistake
Nearly everyone makes this one. Named plainly so you can skip it.
The mistake that looks like success
Harder to catch, because the output still seems fine.
The shortcut that backfires
Usually saves a minute and costs an hour.
The wrong tool entirely
Sometimes the answer is that AI isn't what you needed here.
The Four Kinds of Lesson
Not every lesson teaches the same kind of thing, so the second page changes depending on what you're learning.
The four formats
An idea you need to understand before anything else makes sense. Page 2 covers: How to do it · Key takeaways · How to know you've got it · Common mistakes.
Something you practice until you're good at it. Page 2 covers: How to do it · What good looks like · How to know you've got it · Common mistakes.
Several steps chained into something bigger. Page 2 covers: The steps in order · What you end up with · How to know you've got it · Where it breaks.
Deciding well when there isn't a clean right answer. Page 2 covers: How to think it through · Questions to ask · How to know you've got it · Common traps.
What every lesson shares
Page 1 is always the same shape: the problem it solves, the idea, where you'd use it, three skill levels, related lessons, and an assignment. That means you can compare any two lessons directly even when they teach very different things.
The format is printed in the header next to the lesson number, so you know what kind of thing you're about to learn before you read a word of it.
The Lessons
Short, practical, and built around one thing you can actually try.
Digital vs. Analog
Old computers only ever said yes or no. AI says “probably.” That one change explains almost everything that feels strange about it.
Two ways to think
A record is analog. The groove is a physical wave and the needle traces every wobble in it. Nothing gets rounded off, because there are no steps to round to.
A CD is digital. The same music is measured about forty-four thousand times a second and each measurement is saved as a number. Between two measurements there is nothing. A light switch is the same idea at its simplest: on or off. A dimmer is the record.
Where each one shows up
- Digital: a CD, a calculator, a password check
- Analog: a record, a dimmer, how confident you feel
- AI sits in a strange middle: digital hardware, analog thinking
The bit that surprises people
When AI picks the next word, it is not looking one up. It is scoring thousands of options and giving each a percentage. Maybe “cat” is 41%, “dog” is 22%, “bicycle” is 0.004%. Then it picks, usually from near the top.
That is why the same question can give you a different answer twice. Nothing broke. You are watching a machine make a judgement call instead of a lookup, which is much closer to how you think than how a calculator thinks.
Why it matters
Certainty is not on the menu
A calculator can be exactly right. A model can only be very probably right. Those are different things and need different amounts of checking.
Confidence is a percentage, not a feeling
When it sounds sure, that is a writing style, not a measurement. The actual number is hidden from you.
Small nudges change the odds
Because everything is probabilities, a slightly clearer question really does shift the answer. That is why prompting works at all.
Both kinds are still needed
Use binary tools for anything that must be exact. Use AI for things where a very good guess is genuinely useful.
The Machine That's Always Guessing
You've used it. But do you know what it's doing when it answers you, or why it sometimes makes things up with total confidence?
The idea
A language model doesn't look things up. It predicts the next piece of text based on learned patterns. That explains almost everything about how it behaves.
It's why it writes fluently, why it can be confidently wrong, and why how you ask changes what you get. Understand prediction, and the rest of this book makes sense.
Where you'd use this
- Deciding whether to trust an answer you got
- Explaining why AI got something wrong
- Choosing the right tool instead of defaulting to one
- Any class where you're asked to use AI responsibly
Skill levels
Work up from wherever you areBeginner
Know what it is
- A model predicts text, it does not search
- It has a training cutoff date
- Fluent does not mean correct
Intermediate
Know why it fails
- Why hallucinations happen at all
- What a context window is and why it runs out
- Why the same question gives different answers
- Where training data bias comes from
Advanced
Explain it to others
- Describe tokens and prediction in plain language
- Compare model types and when each fits
- Predict in advance where a model will struggle
Build this · Case 01: The Two-Hour Promise
Riverside Family Clinic hired Bit to answer questions about hours, insurance, and wait times, day and night, alongside their front desk and online booking page. Most days Bit gets it right. Some days he confidently tells a patient something untrue, and nobody notices until they're standing in the waiting room, annoyed.
Step 1, state the problem: in one sentence, what is Riverside's real problem here? Not “the assistant is bad.” Something more specific.
Step 2, pick your goal: more money in (more patients book and show up) or less money out (front desk stops fixing Bit's answers)?
Step 3, build a fix: whatever moves the goal you picked. A better prompt, a short policy on what Bit should never guess at, a sign for the waiting room; the medium is yours to choose.
The Machine That's Always Guessing
Lesson detail
How to do it
Start with a guess
Before reading further, write down how you think AI generates an answer. You'll check this later.
Learn the core loop
Read about next-token prediction. The model picks a likely next piece of text, then repeats.
Find the limits
Identify three things a predictive model cannot reliably do, such as cite a source it never saw.
Test your model
Go back to your guess. Correct whatever was wrong, in your own words.
Key takeaways
It predicts, it doesn't retrieve
Answers are generated from patterns, not looked up in a database. Nothing is being searched unless a tool is explicitly doing that.
Confidence is not accuracy
The model has no internal sense of being unsure. It writes a wrong answer in exactly the same tone as a right one.
Input shapes output
Because it continues patterns, what you give it heavily determines what you get. That's why prompting is a real skill.
How to know you've got it
Can you explain it without jargon?
If you can tell a friend how AI works in two sentences without saying "algorithm," you understand it.
Can you predict a failure?
Name a question you're confident a model will get wrong, then test it. Being right means you understand the limits.
Can you pick the right tool?
You know when AI helps and when a search engine, calculator, or textbook is better.
Can you say what a wrong answer costs?
Name who pays for it: the business loses a customer, or someone burns an hour undoing it.
Common mistakes
Assuming it searched the internet
Unless a tool is explicitly turned on, it isn't looking anything up. It's generating from memory of patterns.
Treating it as a calculator
Arithmetic is not what prediction is good at. Use an actual calculator for actual math.
Believing citations on sight
Made-up sources look exactly like real ones. Every citation needs checking.
Thinking one model is all of AI
Different tools behave very differently. The one you tried first is not the whole field.
The Hidden Cost of Not Defining Your Terms
You asked for “short” and “professional.” You got neither, because you never said what those words meant.
The idea
Every prompt is full of words that feel precise and aren't: short, simple, professional, improve. Each has many meanings, and the model quietly picks one for you.
The cost isn't one bad answer. It's the round trips: you regenerate, you nudge, you settle for close enough, and never notice you accepted a definition you didn't choose.
Where you'd use this
- Any prompt where you had to ask twice to get what you meant
- Group projects where everyone pictured a different deliverable
- Anything where “make it better” was your actual instruction
Skill levels
Work up from wherever you areBeginner
Spot the vague words
- Notice which words in your prompt are undefined
- Know the usual suspects: short, clear, professional
- Catch yourself saying improve without saying how
Intermediate
Define them inline
- Replace vague words with measurable ones
- Give a range, a count, or an example instead of an adjective
- State the audience, which defines tone for you
Advanced
Reuse your definitions
- Keep a short list of terms you define the same way every time
- Write prompts a classmate could run
- Notice when a term needs redefining
Build this · Case 09: Nine Versions, Zero Usable
Ferro Print & Design hired Bit to draft their flyers, menus, and social posts. A client asks for something “clean and professional,” Bit guesses, sends a draft, gets “not quite,” and guesses again. Nine rounds later, everyone's still guessing.
Step 1, state the problem: in one sentence, what is actually costing Ferro here? Be specific about where the process breaks down.
Step 2, pick your goal: more money in (Ferro finishes more paying jobs a week) or less money out (no more unpaid do-overs)?
Step 3, build a fix: whatever moves that goal. An intake form, a brief template, a finished sample; the medium is yours.
The Hidden Cost of Not Defining Your Terms
Lesson detail
How to do it
Write the prompt you'd normally write
Don't improve it yet. You need the honest version to compare against.
Circle the soft words
Any word a reasonable person could read two ways. Most prompts have three or four.
Replace each with something checkable
A word count, a reading level, a named audience, a format, or a short example.
Run both and compare
Same tool, same session. The gap between the two outputs is the cost you'd been paying.
Key takeaways
Adjectives hide decisions
Short, clear, and professional all sound like instructions. They're actually decisions you handed off without noticing.
Numbers and examples beat description
“Under 150 words, for a first-year student” does more work than a paragraph describing the tone you want.
Defined terms are reusable
Once you've decided what professional means for your work, you can paste that definition into every future prompt.
How to know you've got it
Can you name the vague words in your own prompt?
Reading your prompt back and finding three undefined terms is the whole skill.
Do you give examples instead of adjectives?
Showing one line of what you want beats describing it every time.
Could someone else run your prompt?
If a classmate gets your result, your terms were defined.
Did your round trips drop?
First or second try, not the fifth. Each round you skip is money nobody spends.
Common mistakes
Assuming the model shares your context
It doesn't know your course, your professor, or what your class means by “analysis.”
Defining tone with more adjectives
Swapping professional for polished changes nothing. Name the audience instead.
Over-defining everything
You don't need to specify all twelve variables. Define the two or three that actually matter to you.
Never writing it down
Redefining the same terms from scratch every session is the slow version of this mistake.
It Sounds Right. That's the Problem.
The most expensive mistake students make with AI is submitting something they never verified. It reads well, so it feels finished.
The idea
AI output is a draft written by something with no concept of being right. It can invent quotes, misattribute ideas, cite papers that don't exist, and do it all in confident, well-structured prose.
Verification is a skill you can learn and apply quickly. It's also what separates people who use AI well from people who get caught out by it.
Where you'd use this
- Before submitting anything for a grade
- Any time a claim includes a number, date, quote, or source
- Research where you'll cite what you found
- Deciding whether an answer is good enough to act on
Skill levels
Work up from wherever you areBeginner
Spot the risk
- Know which claim types fail most
- Recognize made-up citations
- Notice overconfident language
Intermediate
Verify efficiently
- Check claims against primary sources
- Cross-check with a second tool
- Ask the model to argue against itself
- Track what you verified and what you didn't
Advanced
Verify by default
- Build a routine you run every time
- Judge source quality quickly
- Know when AI shouldn't be used at all
Build this · Case 17: The Newsroom That Got Burned Once
The Weekly Current hired Bit to help reporters draft fast on a packed schedule. Editors are stretched thin, and the newsroom already got burned once by a statistic Bit invented, confidently, in a draft nobody caught in time. Nobody has changed how they work since.
Step 1, state the problem: in one sentence, what is the actual risk The Weekly Current is running by not changing anything?
Step 2, pick your goal: more money in (readers still trust the paper) or less money out (no days lost to corrections)?
Step 3, build a fix: whatever moves that goal. A checklist for editors, a short verification workflow, a one-page policy for reporters; the medium is yours.
It Sounds Right. That's the Problem.
Lesson detail
How to do it
Flag the checkables
Underline every number, date, name, quote, and citation. Those are your verification list.
Go to the source
Find the original. A real paper, a real page, a real book, not another AI summary.
Cross-check
Ask a different tool the same question. Disagreement is a signal to dig further.
Decide and note
Keep what survives, cut what doesn't, and note what you couldn't confirm.
Key takeaways
Specific claims fail most
Numbers, dates, quotes, and citations are the highest-risk parts of any AI answer.
Fluency is not evidence
Well-written and well-structured says nothing about whether it's true.
Verification is fast once it's a habit
Checking five claims takes a few minutes. Submitting a fake citation costs far more.
How to know you've got it
Do you check before submitting?
Verification happens automatically, not only when something feels off.
Have you caught a fake citation?
Once you've found one yourself, you'll never fully trust one again.
Can you tell risky claims from safe ones?
You know a general explanation is safer than a specific statistic.
Can you name what a miss would cost?
A correction burns a day of work, and some of the trust that pays everyone here.
Common mistakes
Trusting it because it sounds academic
Formal tone is a writing style, not a truth signal.
Checking with the same tool
Asking "are you sure?" often just produces a confident repeat. Use a different source.
Verifying only the conclusion
Errors hide in the supporting details, which is exactly what gets cited.
Assuming a real author means a real paper
Models often attach invented titles to genuine researchers.
Glossary
Glossary
continued