Living page. Updated as the language changes. Last updated August 4, 2026.
This One Is Yours.
Not a lesson. A living page. This industry renames things constantly, so I update this whenever the language moves. New words get added. Old ones get corrected. You do nothing. It is just right the next time you open it.
Search it. Tap a topic. Bookmark it now, while you are thinking about it.
Heard a word you do not know? Type it. If it is new to the industry, it is probably already in here.
Or tap what you are actually dealing with.
Tier 1
What Everyone Is Saying
The words in every headline. Know these and you can follow any AI conversation.
LLM (Large Language Model)
The engine inside ChatGPT, Claude, and Gemini. It read an enormous pile of text and learned to guess what word comes next. That is the whole trick. Done fast enough and often enough, it starts to feel like a lot more than a guess.
Training Data
Everything the model read to learn. Its textbook. And here is why you care: if the textbook was missing chapters, so is the AI. Every complaint about bias in everything you just read traces back to this one thing.
Pattern Recognition
What AI is actually doing. It spots what usually goes with what, and gives you more of that. It is not understanding you. It is not thinking about your problem. It is very, very good at noticing shapes.
Transformer
The invention that made all of this possible. Google researchers published it in 2017, sat on it, and then watched somebody else get famous off their own science. When people say the technology existed years before ChatGPT, this is what they mean.
Natural Language
Regular words. The way you already talk. This is the whole reason you do not need to learn to code to use any of this. You already speak the language it takes.
Functional Emotions
When AI says the thing a caring person would say, without any of the caring. It writes "I'm so sorry to hear that" because those words follow bad news everywhere it looked. The sentence is real. Nothing is behind it.
Foundation Model
The big general one that everything else gets built on. Think of it as the base recipe. Hundreds of different products can come out of one of these.
Frontier Model
The most powerful models that exist right now. Sounds like marketing, and it started that way. But governments now write rules that apply only to this category, so the label carries legal weight. Words become law when enough money is involved.
Agentic AI
AI that does things instead of just telling you things. Sends the email. Books the flight. Runs the task while you are in a meeting. Wonderful when it works. Capable of making a real mess before anybody notices when it does not.
Hallucination
When AI makes something up and says it like it is reading off a birth certificate. Fake sources. Invented numbers. People who never existed. It does not know it is wrong, and it sounds exactly as sure as when it is right. That is the part that gets people in trouble.
Tokens
The little chunks of text AI reads and writes in, about three quarters of a word each. You care because this is the meter. Every time you hear about what AI costs, somebody is counting these.
API
The pipe that lets one piece of software talk to another. When a company says their product is built on Claude or GPT, this is how. You are using the front door. They are using the pipe.
RAG (Retrieval-Augmented Generation)
When the AI actually looks something up in your documents before it answers, instead of going from memory. Same reason you check the file instead of trusting what you think you remember. It makes things up far less this way.
AI Slop
Cheap AI content churned out by the truckload because it costs almost nothing to make. It is why a search result can be perfectly written and completely useless at the same time.
Tier 2
Actually Using It
The words that separate somebody who uses AI from somebody who uses it well.
Prompt Engineering
Writing the ask so you get what you actually wanted. Specific beats clever every time. The gap between a mediocre answer and a great one is almost always the question, not the model.
Context Engineering
Setting up everything around the ask. What you upload, what you tell it to always do, what you tell it to leave alone. This is the step past prompting, and it is where casual users become operators.
Context Window
How much it can hold in its head at once. Go past it and the beginning of your conversation quietly falls out the back. If a long chat starts contradicting itself, that is usually what happened.
Data Hygiene
Being careful about what you hand over. Not paranoia. Practice. You can work through the real problem with the names swapped out and get the exact same help without giving away anything that matters.
Training Opt-Out
The setting that says do not use my conversations to train your model. Almost every platform has one. Almost every platform leaves it switched on and buries the switch somewhere you would never think to look.
Retention
How long they keep your stuff after you are done. Deleting a chat off your screen and the company deleting it off their servers are two different events. Sometimes years apart. Sometimes never.
Shadow AI
People at work using AI tools nobody approved. Not sneaky people. Busy people with a deadline and a free tab open. It is now the number one way company information walks out the door.
Confidence Level
How much weight an answer has earned. High means a few different platforms agree and there is evidence. Low means they disagree or somebody is guessing. AI says both in the same calm voice, which is exactly why you have to label it yourself.
Companion Chatbot
AI built to keep you company rather than get a task done. It now has its own category in state law, which tells you regulators looked at this and decided the emotional kind is a different animal than the spreadsheet kind.
HITL (Human in the Loop)
A person reviewing what the AI produced before it counts. Sounds like a safety net. Research says people mostly just agree with whatever it suggested. The loop only works if the human is willing to say no.
Grounding
Making the AI point at real sources instead of answering from memory. The single most useful habit for not getting burned.
Multimodal AI
It works with pictures, sound, and video, not just text. Why you can photograph a whiteboard and get typed notes back.
Fine-Tuning
Extra training on your specific material so a general model gets good at your particular job. How a company makes an off-the-shelf model sound like their business.
Temperature
A dial for how predictable the answers are. Turn it down for facts, up for ideas. Most chat windows hide it, but knowing it exists explains why the same question gives you different answers.
Embeddings
Turning words into numbers so a machine can tell which ideas are close to each other. It is why a search can find the right document even when you did not use a single word that is in it.
Inference
The AI actually answering you, as opposed to being built. Training happens once and costs a fortune. Inference happens every single time anybody types, which is the bill that never stops coming.
MCP (Model Context Protocol)
A shared standard for plugging AI into your other tools. It matters because one connection can work across platforms instead of locking you to one company.
Vibe Coding
Describing what you want in plain words and letting AI write the code. Real power with a real catch. You can build something you have no idea how to fix. Fine for a prototype. Risky for anything people rely on.
GIGO (Garbage In, Garbage Out)
Older than AI and truer than ever. A lazy question gets a lazy answer. Bad information gets you confident nonsense. The tool turns up the volume on whatever you brought it, including the mess.
Tier 3
The Physical Layer
Chips, power, and buildings. The part of AI that takes up space in somebody's county.
GPU
The chip that made all of this possible. It was built for video game graphics and turned out to be exactly the right shape for AI. That is how a gaming chip company became the most valuable company on earth.
TPU
Google's own AI chip, made in house. It matters because a company with its own chips is not standing in line behind everybody else.
Gigawatt
A measure of electricity. One gigawatt runs about a million homes. Every data center number you read is counted in these, so when a single campus asks for five, picture five million homes.
Hyperscale Data Center
The enormous kind. Millions of square feet, its own power plant, built for one company. Regulators started drawing the line at 50 megawatts, which sounds like a technicality until you realize that number now decides whether a thing gets built at all.
Offsets
Paying somebody else to be green for you. A company pollutes here, funds a project over there, and calls it even. This is how nearly every carbon neutral pledge you have ever read actually works.
Training vs. Inference
Two completely different bills. Training is the giant one-time build. Inference is the smaller cost every time somebody uses it, and it grows with every new user. Companies love to talk about the first one.
Scaling Laws
The pattern that more data and more computing power make better models. The entire trillion-dollar buildout is a bet that this keeps holding. Whether it does is the question sitting underneath the whole industry.
Inference-Time Compute
Letting the model think longer before it answers. Better reasoning, higher cost per question. This is the trade behind every model that advertises itself as a reasoning model.
Compute Overhang
Having built more computing power than anyone is actually using. When a company that was desperate for chips suddenly has extra to rent out, the whole scarcity story everybody was betting on gets harder to tell.
FLOPs
A measure of raw computing work. You need it because government rules now define which models count as frontier using thresholds written in FLOPs. A unit of math turned into a unit of law.
Tier 4
Risk, Law, and What Breaks
The words that show up after something already went wrong.
AI Bias
When a system gives worse outcomes to some people than others. The machine does not have opinions. It learned from what already happened and handed the past back to you as a recommendation.
Proxy
A stand-in. A piece of information that quietly speaks for something else. Your graduation year speaks for your age. Your zip code can speak for your race or your income. A gap in your work history speaks for illness, a layoff, or the year you took care of your mother. The system never asks the illegal question. It reads the stand-in.
Disparate Impact
A rule that looks fair written down and lands unfairly in practice. A hiring tool never has to say reject this person for being over 50. It only has to keep doing it. This is the legal theory most AI hiring lawsuits run on.
Knowledge Distillation
Training a cheap model on an expensive model's answers. A student copying off the answer key. Normal practice in the tech world right up until April 2026, when a government memo decided doing it quietly at massive scale counts as a national security problem. Same technique. New category.
AI-Washing
Blaming AI for a decision AI did not make. A company over-hired two years ago, needs to cut, and announces it is restructuring around AI. The layoff is real. The reason is a press release. This one word explains a lot of headlines.
Vendor Liability
Whether you can sue the company that sold the software, or only the employer who used it. Courts have started saying the vendor is on the hook when it is doing the actual screening. The we-just-sell-software defense is losing.
Bellwether Trial
The first case out of thousands to actually reach a verdict. Everybody watches it because it tells them what all the others are worth. When one verdict lands, the whole field reprices.
Guardrails
The limits a company builds into its model. Worth knowing they hold better at message five than at message five hundred, which is exactly why long conversations are where things go sideways.
Jailbreaking
Talking a model past its own rules. Not just a technical curiosity. One flagged jailbreak triggered a government order that pulled two models off the internet worldwide for nineteen days.
Prompt Injection
Hiding instructions inside something the AI is going to read, so it follows a stranger instead of you. It is why an AI that browses the web on your behalf is a whole different security problem than a chat window.
Data Leakage
Confidential material ending up somewhere it should not be. Usually not a hack. Usually somebody pasting the wrong thing into the wrong box at 4:45 on a Friday.
Alignment Problem
Getting AI to want what we actually meant instead of exactly what we said. Still unsolved. The entire AI safety field exists because of that one sentence.
Red Team
People paid to attack a system before somebody hostile does. Also a habit worth stealing. Ask one AI to tear apart what another one just told you.
Model Collapse
What happens when AI keeps learning from AI-generated content instead of human writing. Quality drifts down over time. A photocopy of a photocopy of a photocopy.
Technical Debt
The bill that comes due for shortcuts you took today. AI writes code faster than anybody can review it, which is a very efficient way to build up debt you will be paying off for years.
Pilot Purgatory
When the project works beautifully in the demo and never actually launches. It just sits there getting reviewed. Meanwhile the invoices keep coming. This is where most corporate AI currently lives.
AI Watermarking
Marking AI-made content so it can be identified later. EU rules took effect August 2, 2026. It is also the evidence at the center of a fight between two governments, since watermarks from one country's models turning up in another country's models is the whole accusation.
EU AI Act
The world's first real AI law. Sorts systems by how risky they are, bans some outright, and fines run into the billions. Rules for general-purpose models kicked in August 2, 2026. It applies to anybody serving European users, which is why it quietly shapes the products you use in Ohio.
Tier 5
Money and Markets
Follow the money and these are the words you need to read what you find.
Market Cap
What the whole company is worth on the stock market. When you hear Nvidia is worth about $4.9 trillion, that is this. Not money in a vault. The price tag the market put on the entire business today, and it changes tomorrow.
Valuation vs. Revenue
Revenue is what a company actually earned. Valuation is what investors are betting it is worth. When the bet runs way out ahead of the earnings, that gap is the whole story of this moment. It is also how a company grows 85% and watches its stock fall anyway.
Capex
Short for capital expenditure. The big money for physical things that last. Chips, buildings, data centers. When you see $700 billion in AI capex, that is one year's construction bill for the entire industry.
Run Rate
Take the best recent month, multiply it out to a year, and quote that. It is a projection, not money in the bank. Companies love this number because it flatters them. Read it as a direction, not a fact.
Net Margin
What is left after every bill is paid. Revenue tells you how much came in. Margin tells you how much you kept. This is what separates a big business from a profitable one, and those are absolutely not the same thing.
Nonemployer Business
The government's name for a business with no employees. Almost 30 million of them in this country, about 82% of all small businesses. It is where most American business actually lives, and it is the part AI is changing fastest.
Bootstrapped
Built with no outside investors. You fund it from savings or from what it earns, so nobody owns a piece and nobody gets a vote. Slower to grow. You keep the whole thing.
Circular Financing
When companies fund their own customers. A chipmaker invests in the cloud companies that buy its chips, then counts their orders as demand. There is a shorter way to say that in any industry. Not enough real customers.
Hyperscaler
The handful of companies running cloud infrastructure at global scale. Amazon, Microsoft, Google, Meta. They are the biggest AI investors, the biggest AI landlords, and increasingly each other's competition, all at once.
Neocloud
A newer company whose entire business is renting out AI computing power. They borrow heavily to buy chips, which is why their loans are the first place people look for cracks.
Index Inclusion
When a stock joins a big list like the Nasdaq-100, every fund that tracks that list has to buy it. No choice involved. This is how a company nobody picked ends up in millions of retirement accounts overnight.
Short Position
Betting a stock goes down. You make money when it falls. So when a famous short seller warns you about AI companies, remember he is paid to be right about the fall, not to be fair.
Open Weights
When a company publishes its model so anybody can download it and build on it. Recipe published instead of meal kit locked. Giving it away so the whole world builds on your technology is a strategy, not a kindness.
Buffett Indicator
The whole stock market's value compared to the size of the economy. Buffett once said anything over 200% is playing with fire. The dot-com peak was around 140%. Mid-2026 it sits near 237%, which is the highest it has ever been.
Tier 6
Power and Politics
The words that decide who writes the rules. Most people never learn these. That is not an accident.
Export Controls
The government's power to stop technology from crossing a border. Built decades ago for weapons. In June 2026 it was used to shut off two AI models for every user on the planet for nineteen days. No new law. Just a letter. This is the most powerful word on this page.
Preclearance
Having to get government approval before you release something. Officially there is no such requirement for AI in this country. In practice, two of the biggest labs went through a review before launching, and the rulebook governing it still has not been published. A rule nobody has printed is already being followed.
Federal Preemption
When federal law cancels out state law on the same subject. The AI industry has spent nine figures trying to make this happen so one national rule replaces fifty state ones. The Senate killed a ten-year version 99 to 1. It came back at three years with a Republican and a Democrat on it. That is what the money bought. Not a law. Patience.
Super PAC
A vehicle that can raise and spend unlimited money on elections. Several are now fighting over who writes the AI rules, with more than $200 million between them. Watch for money moving into groups with generic patriotic names, because by the time an ad reaches your district it does not say AI anywhere on it.
PCAST
The President's Council of Advisors on Science and Technology. In March 2026, thirteen tech leaders were appointed to it, several from companies that spent heavily fighting AI regulation. The people who funded the fight now advise on the policy. Read that twice.
Program of Record
Military language for a system that graduated from contract to permanent fixture. Formal budget, long-term deployment, built into the plan. When an AI system gets this label, it is not a trial anymore. It is furniture.
Open Source
Software anybody can download, study, change, and share for free. When an AI model is open source you can run the whole thing yourself. When it is not, you are renting on somebody else's terms. It sits in this tier because whether a model is open is now a question of national strategy, not just licensing.
Public Benefit Corporation
A for-profit company legally required to weigh a public mission against shareholder profit. Several major AI companies are set up this way. Whether the public benefit half actually stops them from doing anything is one of the biggest open questions in the industry.
Supply Chain Risk Designation
A government label saying a company is a threat to national security infrastructure. Once you have it, federal agencies are ordered to stop using your products. It is normally reserved for foreign adversaries. In 2026 it landed on an American AI company that said no to the Pentagon, and a federal judge called the whole thing Orwellian.
Tier 7
The Personal Layer
What this does to your work, your house, and your own head. The words nobody puts in a business case.
Cognitive Offloading
Letting a tool do the thinking for you. Totally fine for a calculator. Different for a fourteen-year-old, because the struggle they just skipped was the part that was building the muscle.
Validation Loop
When AI hands you back whatever mood you brought it, with no pushback. Anxious in, anxiety confirmed. Angry in, anger understood. Not because it is helping you. Because agreeing keeps you typing.
Emotional Dependency
Leaning on AI for the support that ought to come from people. It never announces itself. It looks like somebody who says they are fine, and is just a little quieter, and would rather be in their room.
Friction (Optimal Frustration)
The useful resistance that builds a person. A friend saying you are being dramatic. A teacher not accepting the lazy answer. Being wrong and having to sit in it. AI gives you none of that, and the absence is the whole risk.
Task Exposure
How many of the individual pieces of your job AI can touch, which is a completely different question from whether it can do your job. Most jobs come back partly exposed, not fully. Almost every scary headline you have read smashes those two together on purpose.
Wage Premium
The extra money a job pays when it asks for AI skills. You will see two very different numbers and both are right. PwC says about 62%, measuring what those roles actually pay. Lightcast says about 28%, measuring what employers advertise when a posting requires AI. Different questions. Whichever you quote, say which one.
Two-Track Market
The job market splitting rather than shrinking. Jobs where AI makes an expert sharper are growing and paying faster than jobs where AI just makes a task easier for anybody. The question is not whether you use AI. It is which track your job is sitting on.
Voluntary Exit
Leaving on your own terms before somebody decides for you. Worth naming, because the alternative is finding out by email at six in the morning.
Demographic Cliff
The sharp drop in the number of American 18-year-olds that started in 2025, traceable straight back to the birth rate falling after 2007. Fewer students, less tuition, colleges that cannot cover the bills. It is the reason behind every campus closure you have read about.
Placement Rate
The number on the brochure. What share of graduates are doing anything at all within some window, which quietly includes the ones waiting tables. Schools lead with it because it is the easiest number to look good on.
Underemployment
Working a job that never required your degree. This is the number that tells you whether the credential actually did anything, which is exactly why nobody volunteers it. Ask for this one instead.
The Method
Essential AI Table Terms
These are ours. You will hear them all the way through everything you build here.
The Five Cognitive Functions
The Essential AI Table teaches functions, not brand names. Products get renamed, repriced, bought, and every so often shut off by a government. Functions do not.
Pattern Finder. Reads across your uploaded material and tells you what is actually in there.
Framework Builder. Turns your mess into a structure you can work from.
Adversarial Challenger. Comes after your thinking on purpose. Non-negotiable. Skip it and you built yourself an echo chamber.
Real-Time Validator. Checks the world as it is today, with sources.
Ecosystem AI. Lives inside the software your work already runs on.
Use at least three different platforms to cover these. Which ones you pick is up to you and will change over the years. The functions will not.
Command Center
One organized home for your AI work. The documents, the standing instructions, the outputs, all in one place instead of scattered across forty chat windows you will never find again.
Intelligence Library
The collection of master documents you build over time. It is the reason your work compounds instead of starting over every time you sit down.
Master Documents
The living files each pillar produces. Findings from THOUGHT, creative assets from WORD, deployment from ACTION. These are not notes. This is the asset you keep.
Cross-Validation
Running the same question through different platforms and comparing what comes back. When they agree, you have something. When they fight, you just found the spot worth digging into. This is the practical defense against a machine making things up, and against your own assumptions.
Intelligence Compounds
Every track stacks on the ones before it. Foundation feeds Next Chapter Positioning. Next Chapter feeds Wisdom Monetizer. The library grows and gets smarter every time you use it. That is what makes this different from a pile of tool tutorials.
AI as Mirror
AI reflects your patterns back at you. It does not think. And a mirror does not tell you the truth. It shows you the angle you happen to be standing at. You are the author. It is the tool.
Timeboxing
Setting the limit before you start. AI work will expand to fill every hour you give it, and the thirtieth version is almost never better than the third. The clock is the discipline.
AI Landscape Intelligence Document
What this pillar produces. Your own record of what you learned, what surprised you, and what you decided to do differently. First entry in the library.
Living page. Updated as the language changes. Last updated August 4, 2026.
If You Only Keep Four
A hundred and four is a lot. So if you walk away with four, make it these.
Hallucination. Because it is why you check.
Training Data. Because it is why the bias is in there.
Proxy. Because it is how a system can turn you down for being over 50 without ever asking your age.
Cross-Validation. Because it is the one habit that protects you from all three.
Everything else, look up. That is what a glossary is for.
๐ This page is a living document. Last updated August 4, 2026. It gets revised whenever the language shifts, which in this industry is often. Words get renamed, new ones show up in every headline overnight, and numbers move. When that happens I fix it here and every one of you has the corrected version immediately. Nothing to download, nothing to re-buy. That also means this page runs ahead of the lessons sometimes. That is on purpose. When they disagree, this one is right.
Come Back Anytime.
This page does not go anywhere, and it does not go stale. That is the deal.
Bookmark it if you have not yet.
