netvocab.com

APR

2026-09-26T00:00:00.000Z

What it means

APR, short for annual percentage rate, is the yearly cost of borrowing money expressed as a single percentage, including the interest and any fees you cannot avoid.

An interest rate only tells you what the lender charges on what you still owe. Real borrowing comes with extras: an arrangement fee, a product fee, sometimes an account charge. APR folds all of that into one figure, spread across the life of the loan, so deals built differently can be compared.

The closest everyday comparison is a supermarket shelf label. One bottle holds 500ml and the next two litres, so the sticker prices tell you little. The price per litre underneath is the standardised number that makes comparison possible. APR does that job for credit.

Any standardised number has to assume things. APR assumes you keep the borrowing for its full term and pay on schedule. Change either and your real cost drifts from the advertised figure.

Where it came from

The usual account is that APR exists because lenders once quoted prices in whichever way flattered them most. One advertised a monthly rate, another a flat rate on the original amount, a third a sum of interest with no rate at all. All three could describe the same deal, and none could be compared without doing the maths.

In the United States, the Truth in Lending Act of 1968 made lenders disclose a finance charge and an annual percentage rate worked out to a set formula, under the rules known as Regulation Z. The United Kingdom followed with the Consumer Credit Act 1974, which put the annual percentage rate of charge into adverts and agreements. European rules took a similar line, and the mortgage version is labelled APRC.

Regulators kept patching the idea as lenders found the gaps. A headline APR could be quoted for a deal almost nobody was offered, so UK rules brought in the representative APR: the advertised figure must be the rate at least 51 per cent of accepted applicants receive. Several similar-looking percentages are now in circulation, and they are not interchangeable.

Figure	What it covers	Where you meet it
Interest rate	The charge on what you owe, fees excluded	Statements and rate tables
APR	Interest plus compulsory fees, over the full term	Loan and car finance adverts
Representative APR	The APR most accepted applicants must get	UK credit adverts
AER	The savings rate, with compounding included	Savings accounts, not borrowing
APRC	The mortgage version, including the reversion rate	UK and EU mortgage paperwork

How people actually use it

“Representative 24.9% APR variable.” The standard line under a card advert. Variable means it can move, representative means plenty of applicants get worse.
“Borrow £10,000 over five years, 7.9% APR representative.” Loan advertising. The term matters as much as the rate: the same APR over longer means more interest in total.
“0% APR on purchases for 18 months, then 22.9% variable.” A card offer. Free for a window, then priced like any other.
“The payment looks fine, but what’s the APR and how long does it run?” Car finance talk. Low payments often just mean a long contract.
“They quoted a flat rate of 4%. Ask what that is as an APR.” Flat rates are worked out on the original sum, so they sound cheaper than they are.

In a sentence

Dan: This loan is 6.4% and the other is 6.9%, so the first is cheaper.
Priya: Check the APRs. The first has a £400 arrangement fee, and that shows up in the APR but not the rate.

Lena: Zero per cent for two years sounds too good to be true.
Sam: It’s real. The APR after that, and the date it kicks in, are the bits to write down.

Common misconceptions

“APR and the interest rate are the same thing.” They match only when there are no compulsory fees. Add a product or arrangement fee and the APR rises above the rate, which is why it exists.
“Representative APR is what I’ll be charged.” It is the rate a majority of accepted applicants must receive, not a promise to you. Your own quote follows a credit check, and many people are offered more.
“The lowest APR is always the cheapest deal.” Over a short term, often not. APR spreads fixed fees across the whole agreement, so a low APR on a long loan can cost far more in total than a higher APR repaid quickly.
“APR tells me what my credit card costs.” Only if you carry a balance. Clear the statement in full and the purchase APR never bites, though cash withdrawals usually charge interest from day one.

Interest rate: the charge applied to the balance you still owe, before fees.
Representative APR: the advertised APR that a set share of accepted applicants must actually be offered.
AER: the annual equivalent rate, the savings-side figure that includes compounding.
APRC: the mortgage figure, covering the full term rather than the introductory deal.
Flat rate: interest quoted on the original amount borrowed, which flatters the real cost because your balance falls as you repay.
Total amount repayable: the plain sum of everything you hand over, often more telling than any percentage.

Questions people ask

What is the difference between APR and the interest rate?

The interest rate is the charge on your outstanding balance. The APR adds compulsory fees and spreads them over the term, so it is usually higher and the fairer basis for comparing lenders.

Does the lowest APR always mean the cheapest borrowing?

No. APR assumes you keep the agreement for its full term, so a low APR stretched over years can cost more in total than a higher one cleared quickly. Check the total amount repayable too.

Advertised rates are usually representative, so they only have to go to a majority of accepted applicants. Everyone else is priced on their credit file, income and the lender’s rules.

Does 0% APR mean the credit costs nothing?

For the promotional window, usually yes on purchases, but check for a transfer fee and note when the offer ends. After that the standard APR applies to whatever is left.

Is APR worked out the same way everywhere?

The idea is shared, but the formula, the fees counted and the advertising rules differ by country and product. Two similar deals can carry different APRs in different markets.

This is a plain-English explanation, not financial advice, and it cannot tell you which deal suits you. The rules behind APR differ between countries and products, so read the paperwork or ask a regulated adviser before signing. Other money words with unhelpful names, like the escrow account holding part of a US mortgage payment, get the same treatment across the dictionary.

The short version

APR is the yearly cost of borrowing as one percentage, with compulsory fees folded in, so deals built differently can be compared. It is not the interest rate, a representative APR is not your quote, and a low APR over a long term can still cost more than a higher one repaid fast.

Escrow

2026-09-25T00:00:00.000Z

What it means

Escrow is an arrangement where a neutral third party holds money, documents or both until each side has done what it promised, and only then passes them on.

The problem it solves is who goes first. A seller will not sign the property over before the money arrives. A buyer will not send the money before the property is theirs. Escrow closes that gap by parking the valuable thing with someone who gains nothing either way.

Picture two children swapping toys, each eyeing the other suspiciously. Hand both toys to a parent, who returns them in opposite directions at the same moment, and neither child has to trust the other. They only have to trust the parent.

The word takes several shapes. Something is held in escrow, a deal is in escrow, and lawyers say funds were escrowed. In the United States it also names the account your mortgage servicer keeps for property tax and insurance, topped up monthly alongside the interest your APR describes.

Where it came from

The usual account traces the word to Old French escroue, a scrap or roll of parchment, from the root that gave English scroll and shred. In older English law, a deed could be handed to a third party to pass on once a condition was met, and while it waited it was said to be delivered as an escrow. The word described the document first. The money came later.

The arrangement exists because some deals cannot happen in one movement. Buying bread is simultaneous: cash out, loaf in. Buying a house involves searches, a lender, a land registry and signatures gathered over days, so there is always a stretch where one side has committed and the other has not.

Escrow fills that stretch with a third party bound by instructions agreed in advance. Who that party is varies by country. Much of the United States uses licensed escrow or title companies. In England and Wales the money sits in a solicitor’s client account, regulated but rarely called escrow. Parts of Europe use a notary, and online the holder is often the platform itself.

Arrangement	Who holds the money	What releases it
Escrow or title company	A licensed third party firm	The written conditions are met
Solicitor’s client account	The conveyancing solicitor	Completion of the sale
US mortgage escrow account	Your loan servicer	Tax and insurance bills fall due
Marketplace payment hold	The platform	Buyer confirms delivery, or a window closes
Tenancy deposit scheme	A government-backed scheme	End of tenancy, less agreed deductions

How people actually use it

“The house has been in escrow for three weeks.” An American seller. It means under contract with conditions outstanding, not sold.
“We’ll release the funds once the domain shows in your account.” A domain sale, where both sides are strangers and the asset moves instantly.
“Part of your payment goes into escrow for property taxes and insurance.” A US mortgage statement. Here escrow means a pot the servicer manages for you.
“The client will put the first milestone in escrow before we start.” Freelance work. The money is committed and visible, but not yet yours.
“Your deposit is held by our solicitor until completion.” UK conveyancing. Same function, different vocabulary.
“Escrow fell through after the survey.” A deal that collapsed while conditions were still being checked.

In a sentence

Aisha: He wants the code before he pays, and I want paying before I hand over the code.
Tom: Put it in escrow. The platform holds the money, you upload, it pays out.

Ben: So the house is ours?
Sara: Not yet. It’s in escrow. The survey and the searches still have to come back clean.

Marc: My mortgage payment went up but the rate hasn’t changed.
Jo: Check the escrow portion. Your council’s tax bill or your insurance premium probably rose.

Common misconceptions

“The escrow agent is looking after the buyer.” They are not on anyone’s side. An escrow agent follows the written instructions both parties agreed, which is the point. If those instructions favour the seller, the agent will carry them out anyway.
“Money in escrow is automatically safe.” Escrow guards against one risk: the other side taking your money or goods and vanishing. It does not rescue you from a poorly drafted contract, an agent who becomes insolvent, or a fraudster emailing fake payment details that look like the solicitor’s. That last one is common enough that firms now tell clients to phone and confirm bank details.
“In escrow means the sale is done.” It means the sale is under way with conditions still open. Deals in escrow fall apart regularly.
“Any trusted friend can act as escrow agent.” For anything sizeable the holder is licensed or regulated, insured, and required to keep client money apart from its own. That structure is most of the protection.

Escrow agent: the neutral person or firm holding the money and following the instructions.
Client account: a regulated account where a law firm keeps money belonging to clients, separate from its own.
Completion, or closing: the moment the deal finishes and money and title change hands.
Contingency: a condition, such as a survey or a mortgage offer, that must be satisfied before release.
Earnest money: in US property, the upfront sum a buyer puts in escrow to show they are serious.
Title company: an American firm that checks ownership records and often holds the escrow.

Questions people ask

Who pays the escrow fee?

It depends on the deal and local custom. In US property sales the fee is often split or negotiated, and on freelance and domain platforms it usually comes out of the seller’s proceeds.

Is escrow the same as a deposit?

No. A deposit is money you hand to the other party, who holds it. Escrow money goes to a neutral third party, so neither side can spend it or withhold it on a whim.

How long does money stay in escrow?

As long as the conditions take. A domain transfer may clear in days, a property sale commonly runs for weeks, and a US mortgage escrow account lasts the life of the loan.

Can money in escrow be returned?

Yes, if the conditions are not met or both sides instruct the agent to unwind the deal. Where the two disagree about who should get it, the agent may hold the money until they settle or a court decides.

Does the United Kingdom use escrow?

The arrangement is everywhere in UK conveyancing, but is usually described as money held in a solicitor’s client account until completion. The word escrow appears more in commercial contracts and software deals, such as source code escrow.

This is a plain-English explanation of a word, not legal or financial advice about your own transaction. Who may hold escrow money, how it is protected and what happens in a dispute differ by country, so ask a qualified professional where the deal is taking place, and browse the dictionary for the rest of the paperwork vocabulary.

The short version

Escrow is the neutral middle ground of a deal: a third party holds the money or documents until the agreed conditions are met, then releases them. It settles the awkward question of who goes first, but it is no guarantee that the deal is sound or that nobody can be defrauded.

AI jargon, translated

2026-09-13T00:00:00.000Z

The AI boom arrived with a vocabulary nobody handed out. One month the news was about chatbots. The next it was tokens, agents and context windows, used by people who assumed everyone already knew. Product pages are worse. They fit three of these terms into a single sentence and move on.

Most of the words are simpler than they sound, and they get simpler again once you sort them by what they describe. So this glossary is grouped by job, not alphabet: the basic ideas, how models learn, how you use them, how people build with them, what goes wrong, and the slang that grew up around it all.

The basics

These are the words everything else rests on, and the ones used most loosely. A lot of confusing AI coverage comes from treating them as synonyms when they are really nested. Machine learning is one way of doing AI. A large language model is one kind of machine learning model. The chatbot on your screen is a product wrapped around that model, with an interface and some rules added.

Term	Plain English	Why it matters
AI	Artificial intelligence. Computer systems that do tasks we used to assume needed a human mind, such as recognising faces, translating or writing.	It is an umbrella term. A spam filter and a chatbot both qualify, so “AI-powered” on a label could mean almost anything.
Machine learning	Software that learns patterns from examples instead of following rules a person wrote out by hand.	Most of what gets called AI today is machine learning underneath.
Model	The trained system itself: the thing that takes an input and produces an output.	The same model can sit behind many different products, so two apps with different names may be running the same engine.
LLM	Large language model. A model trained on vast amounts of text to predict what comes next, which turns out to be enough to write, summarise and hold a conversation.	It is the technology behind most of the chatbots people mean when they say “AI” now.
Generative AI	AI that produces new material, such as text, images, audio, video or code, rather than sorting or scoring what already exists.	It is the branch behind the current boom, and the one raising the loudest questions about copyright and work.
Training data	The examples a model learns from. For language models that means enormous amounts of text, much of it gathered from the public web.	A model can only reflect what it was shown, which is why arguments about bias and copyright usually start here.
Parameters	The internal numbers a model adjusts as it learns. Together they hold what it has learned, in a form no person can read directly.	Model size is quoted in parameters, often in the billions. More usually means more capable and more expensive to run, but not automatically better at your task.

How models learn

A model is not written line by line like ordinary software. It is grown from examples, in stages, on a great deal of expensive hardware. This is the part of the story that makes the business pages, because it is where the money, the electricity and the arguments over whose data was used all come in.

Term	Plain English	Why it matters
Training	Showing a model huge numbers of examples and nudging its parameters until its outputs improve.	It happens before you ever use the model. Your chat does not change the model in real time, though some companies use conversations to train later versions.
Pre-training	The first and largest stage of training, where a model soaks up general patterns from a vast pile of text, images or both.	This is where most of the cost goes. The “P” in GPT stands for pre-trained.
Fine-tuning	Extra training on a smaller, focused set of examples to adapt a model to a task, a subject or a tone of voice.	It is how a general model becomes a legal drafting tool or a support assistant without starting from scratch.
RLHF	Reinforcement learning from human feedback. People rate a model’s answers, and the model is trained towards the kind they preferred.	It is a big reason chatbots sound polite rather than like raw autocomplete. It can also nudge them towards telling people what they want to hear.
Benchmark	A standard test used to compare models, such as a set of maths problems, coding tasks or exam questions.	Launch posts lean on them heavily. A model can score well on a test without being better at your actual work, and test questions sometimes leak into training data.
Compute	The raw processing power used to train and run models. Used as a noun, as in “we need more compute”.	Access to it decides who can build the largest models. In practice it means chips, data centres and very large electricity bills.
GPU	Graphics processing unit. A chip first built for video game graphics that turns out to be very good at running the huge numbers of simple calculations AI needs, all at once.	These chips are scarce and costly, which is how their best-known maker became one of the most valuable companies in the world.

How you use them

This is the vocabulary of the chat box itself. None of it needs technical knowledge, and a handful of these terms explain most of the odd things chatbots do: why they lose track of the start of a long conversation, why the same model behaves differently in different apps, and why pricing is counted in something other than words.

Term	Plain English	Why it matters
Prompt	Whatever you give a model to respond to: a question, an instruction, a document, a photo.	The model works only from the prompt and what it learned in training. Vague in, vague out.
Prompt engineering	Writing prompts deliberately to get better results, by adding context, examples and a clear format for the answer.	Less mysterious than the name suggests. It is mostly the old skill of writing a good brief.
Token	The chunk of text a model actually reads and writes. Often part of a word. In English, a token averages roughly three quarters of a word.	Prices, speed limits and memory limits are all counted in tokens rather than words.
Context window	The amount of text a model can take into account at once, including your messages, any files and its own replies.	Once a conversation outgrows it, earlier parts may be dropped or summarised, which is one reason long chats forget how they began.
System prompt	Instructions a company or developer gives a model before your conversation starts, setting its role, tone and rules.	It explains why one model can act quite differently in two apps. You usually cannot see it.
Multimodal	Able to handle more than one kind of input or output, such as text, images, audio and video.	It is why you can photograph a menu in another language, or a boiler error code, and ask what it means. Quality varies by type.
Inference	Running a trained model to get an answer. Every message you send is a round of inference.	Training is paid for once per model. Inference is paid for every time anyone uses it, so running costs grow with popularity.

Building with AI

These are the words you hear when a company says it is “adding AI” to a product. Relatively few companies train their own models from scratch. Most rent access to someone else’s and build around it: connecting it to their own documents, giving it tools to use, and fencing it in. The terms below describe that plumbing.

Term	Plain English	Why it matters
API	Application programming interface. A standard way for one piece of software to request things from another. AI providers sell access to their models this way.	Plenty of “AI-powered” apps are ordinary software passing your text to another company’s model through an API. Worth knowing when you wonder where your data goes.
RAG	Retrieval-augmented generation. Before answering, the system searches a set of documents and hands the relevant passages to the model along with your question.	It lets a model answer from company files or recent news without retraining, and makes it easier to show sources.
Embeddings	A way of turning text, images or other data into lists of numbers, so that things with similar meanings end up close together.	They power search that matches meaning rather than exact words, and they sit underneath most RAG systems.
AI agent	An AI system that works through a task in steps, using tools such as search, a browser or code, and deciding for itself what to do next.	The word is stretched thin in marketing. Ask what it can do without you, and what happens when it gets a step wrong.
Open-weight model	A model whose trained parameters are published, so anyone can download it and run it on their own machines.	It lets organisations keep data in-house. It is often called “open source”, but the training data and code usually stay private, so that label is disputed.
Guardrails	Rules and filters around a model that stop it producing certain content or taking certain actions.	They are why a chatbot refuses some requests. Set too loose, they let harm through. Set too tight, they block harmless questions.

When it goes wrong

AI has an unusually vivid set of names for its failures. Some describe flaws in the model itself. Others describe ways people misuse it or attack it. Telling them apart helps you judge how worried to be about a given headline, and who is actually at fault.

Term	Plain English	Why it matters
Hallucination	When a model states something false with full confidence, such as an invented quote, statistic or court case.	Models generate plausible text, not checked facts. Verify anything that matters before you repeat it.
Bias	When a model’s outputs treat groups of people unfairly or repeat skewed assumptions, usually inherited from its training data.	It matters most when AI helps make decisions about jobs, loans or policing, where a small skew repeated at scale becomes a pattern.
Jailbreak	A prompt designed to talk a model out of its own rules, often through role-play or an elaborate hypothetical.	It shows that guardrails are fences, not walls. Companies patch known tricks and people find new ones.
Prompt injection	Hiding instructions inside content a model will read, such as a web page, email or document, so it obeys them instead of the user.	It is a serious risk for AI agents. An assistant that reads your inbox can be steered by whoever writes to you.
Deepfake	Realistic fake video, audio or images of a real person, made with AI.	Cloned voices already turn up in phone scams. A convincing clip is no longer proof on its own.
Alignment	The work of making AI systems reliably do what people intend, and not pursue goals or habits nobody asked for.	It covers everything from a chatbot that flatters too much to worries about far more capable systems in future.

The culture and the slang

Any technology used this widely grows a slang layer, and AI’s grew fast. These words move the way we traced in our piece on how new slang spreads: a small group uses a term precisely, bigger accounts carry it out, and it loosens as it travels. A few are neutral. Several are plainly insults, which says something about the mood.

Term	Plain English	Why it matters
Vibe coding	Building software by describing what you want to an AI and accepting the code it writes, often without reading it closely.	It lets non-programmers make working apps. It is also a quick way to ship bugs and security holes nobody understands.
AI slop	Low-effort AI-generated content made in bulk: strange images, filler articles, fake reviews, spam videos.	It clogs search results and feeds. The word “slop” carries the verdict: made cheaply, for volume, with no one checking.
Clanker	A mocking word for robots and AI, borrowed from Star Wars, where clone troopers use it for battle droids.	It is mostly a joke, but it gives a name to real irritation at AI turning up everywhere uninvited.
AGI	Artificial general intelligence. A hypothetical AI able to do most intellectual work as well as a person.	There is no agreed definition or test, so claims that it is close, or already here, are hard to check.
Doomer	Someone who expects advanced AI to end very badly, possibly for humanity as a whole.	Critics use it as a put-down. Some of the worries behind it are taken seriously by researchers, so the label can hide a real argument.
Accelerationist	Someone who wants AI developed as fast as possible and sees speed as the route to progress. The online movement version calls itself “e/acc”, short for effective accelerationism.	It is the opposite camp to the doomers. Most real debates about regulating AI happen in the wide space between the two.
“ChatGPT it”	Asking a chatbot instead of searching, as in “just ChatGPT it”. A brand name turned verb, the way “Google it” was.	It marks a real change in habits. A chatbot answer can sound certain without being checked, so it deserves the same doubt as any single source.

Three terms people mix up

AI, machine learning and LLM. Picture three boxes, one inside the next. AI is the outer box: any system doing something that looks like thinking. Machine learning sits inside it and covers systems that learn from examples rather than hand-written rules. An LLM sits inside that, as one kind of machine learning model built for language. So every LLM is AI, but most AI is not an LLM. The filter catching your spam counts as AI, and it has never chatted with anyone.

Training and fine-tuning. Training builds a model’s abilities in the first place, and the main stage can take months and cost enormous sums. Fine-tuning is a short extra course afterwards, on a narrow set of examples. Training teaches a model language. Fine-tuning teaches it house style. When a company says it has “trained its own AI”, ask which one it means, because fine-tuning an existing model is far more common.

RAG and fine-tuning. Both make a general model useful for a specific job, but they fix different problems. Fine-tuning changes the model itself, so it suits teaching a style, a format or a type of task. RAG leaves the model alone and hands it the right documents at the moment you ask, so it suits facts that change or that live in your own files. A rough rule: fine-tune for how it should answer, and use RAG for what it should know.

How to keep up without reading every launch post

Learn the concept, not the brand name. Model names change every few months and tell you little. The ideas underneath, like context windows, agents and fine-tuning, stay useful for years. It is the approach behind the full dictionary, slang and tech alike: understand what a word does, and the next one doing the same job will not throw you.
Turn every claim into a plain question. “Agentic” means “what can it do without me?” “Multimodal” means “can it look at my photo?” “Longer context” means “can it read the whole report?”
Treat benchmark scores as a trailer, not a review. Try the tool on one real task of your own. Ten minutes of that tells you more than a chart.
Notice when a word stops meaning anything. Once every product is an “agent” or “AI-powered”, the label has stopped carrying information. Ask what the thing does on an ordinary Tuesday.

Questions people ask

What is the difference between AI and ChatGPT?

AI is the whole field, covering everything from spam filters to image generators. ChatGPT is one product within it: a chatbot made by OpenAI and built on large language models.

Why do AI chatbots make things up?

AI chatbots make things up because language models are built to produce likely-sounding text rather than to check facts, so confident wrong answers come from the same process as confident right ones. Connecting a model to real documents reduces the problem without removing it.

What does “tokens” mean on an AI pricing page?

On an AI pricing page, a token is a small chunk of text, often part of a word, and the unit a model reads and writes in. Providers usually quote prices per million tokens and charge for both the text you send and the text the model sends back.

Is prompt engineering a real job?

Some companies have hired prompt engineers under that title, but for most people prompt engineering is a skill rather than a job. It mostly means writing clear instructions with the right context and examples, which is useful well beyond AI.

Is an AI agent just a chatbot with a new name?

An AI agent differs from a chatbot because it carries out tasks in several steps, using tools such as a browser or your files, instead of only replying. The label is applied generously in marketing, so check what a product actually does before trusting the word.

How AI is quietly changing the way we text

2026-09-12T00:00:00.000Z

Your friend has had a rough week. You send a clumsy message to check in. Twenty minutes later the reply arrives, and it is beautiful. It thanks you for reaching out. It lists three things they are grateful for. It closes by hoping you are looking after yourself too.

It is a lovely message. And for a split second, before you feel any warmth at all, you wonder whether they actually wrote it.

The autocomplete era

Machines have been finishing our sentences for longer than most people realise. Old keypad phones guessed your word from a few button presses, often wrongly. Autocorrect made confident mistakes of its own. Then, about a decade ago, email apps began offering tiny ready-made replies. “Sounds good!” “Thanks for letting me know.”

Those buttons were a small convenience. They were also a quiet nudge. Tap one and you have said what countless other people said that day, in the same words. When the easy option is always there, plenty of people take it.

Keyboards now suggest the next word as you type. Some offer to rewrite a whole message in a friendlier tone. Each suggestion is reasonable on its own. Together, they pull everyone towards the same middle: polite, upbeat and slightly bland. The rough edges that made your texts sound like you get sanded down a little at a time.

That is the flattening of tone. You would never spot it in one message. You notice it when everybody’s “congrats!!” starts to look the same.

“Did ChatGPT write this?”

Once chatbots that write whole paragraphs became free and easy to use, a new question entered the group chat. Did a person write this, or a machine?

Tools like ChatGPT are built on an LLM, a large language model trained on vast amounts of human writing to predict the next word. That is why the output sounds so fluent. It is also why it sounds so familiar.

People have built up a folk list of tells. Over-polished structure, with a greeting, three tidy points and a warm sign-off. “I hope this message finds you well.” Lists that always come in threes. And, most famously, the em dash, the long line some writers use to break off a thought. Somewhere along the way it became a supposed giveaway.

Here is the problem. Humans wrote like this first. The chatbots learned it from us. “I hope this message finds you well” has been clogging office inboxes for decades. The rule of three is as old as speechwriting. Novelists and journalists have loved the em dash for centuries. Some now say they cut it from their own writing to avoid suspicion, which is a strange thing to have to do.

The tell people point to	What it usually means
“I hope this message finds you well”	Someone who learned to write emails in an office
An em dash in the middle of a sentence	A person who likes punctuation, as writers always have
A neat list of exactly three things	One of the oldest habits in public speaking
Words like “crucial” and “furthermore”	School essays or years of formal work writing
A long, well organised reply to a short text	They care, or they had time on the train

No single tell proves anything. Sometimes the polished message really was drafted by a chatbot. Often it came from a careful person now mistaken for one. The detective work says more about our suspicion than about the message.

Writing messy on purpose

So people have adapted. If polish looks artificial, mess looks human.

All-lowercase messages. Missing full stops. Typos left in because fixing them feels try-hard. None of this is new. What is new is that some people now do it on purpose, as a kind of proof of life.

Shorthand plays a big part. A word like lwk, short for low-key, or icl, short for “I can’t lie”, does something a chatbot’s default voice does not. It signals that you are talking casually, to someone you know, without trying to impress. “icl that film was mid” reads like a person because it is imperfect. It skips words. It assumes shared context.

Chatbots can imitate this if asked. But their default is tidy and complete, and most people never change it. So for now, a dropped capital letter says, roughly, “a human typed this with their thumbs, probably while doing something else.”

The irony is neat. Adults spent years complaining that texting abbreviations were ruining proper writing. Now they are one of the clearest signs of a real person on the other end.

Ghostwritten feelings

Emotional writing is trickier. Here, AI help raises questions that spelling suggestions never did.

Dating apps are the obvious example. Some offer to help write your bio or an opening line, and standalone apps promise rizz on demand. The appeal is clear, since messaging a stranger is intimidating. The risk is just as clear. If the witty person in the chat was partly a chatbot, the first date means meeting someone slightly different. For plenty of people, finding out a match’s charming messages were machine-made would be an instant ick.

Then there are the harder messages. Apologies. Condolences. Breakups. People ask chatbots for help with these precisely because they matter and hurt to write. Someone who freezes when a friend’s parent dies might want a starting point. Someone ending a relationship might want help being kind rather than blunt.

Does it count if a machine drafted it? It probably depends on what happened next. A message you drafted with help, then rewrote and meant every line of, is not so different from asking a friend how to phrase something. People have always borrowed words from sympathy cards and wiser friends. A message pasted in without reading closely is a different thing. The words may be fine. The effort, often the thing the other person wants most, is missing.

One useful test is simple. Would you be comfortable if they knew? If yes, the tool was probably a helper. If not, it may have done a job that was meant to be yours.

The slang people invented about AI

Whenever something big arrives, people give it nicknames. Naming a thing helps us get a grip on it. Mocking it a little makes it less overwhelming. AI has produced a whole vocabulary very quickly, following the usual pattern of how new slang spreads: small communities first, then everyone.

Clanker is the rudest of the bunch. It started as a Star Wars insult for battle droids and became a dismissive word for robots and AI in general. Calling a customer service bot a clanker is a tiny act of rebellion, and usually a joke.

AI slop describes the flood of low-effort machine-made content filling feeds. Odd images. Recycled articles. People in pictures with too many fingers. The word suggests something cheap, mass-produced and not quite fit to consume.

Vibe coding is the gentler one. It means building software by describing what you want to an AI and accepting whatever comes back, without really understanding the code. The AI researcher Andrej Karpathy coined it in early 2025. It caught on because so many people recognised themselves in it.

“Just ChatGPT it” has joined “just Google it” as the modern way of telling someone to stop asking you things. And people treat chatbots like that one overconfident friend. It answers everything with total certainty. It is sometimes completely wrong. In technical terms, a model can hallucinate, inventing facts, quotes or sources that do not exist. People have also noticed that chatbots love to yap. Ask a yes-or-no question and you may get four paragraphs and a summary.

None of this vocabulary is neutral. It is sceptical, playful and a bit defensive. That is how people sound when adjusting to a big change in real time.

What stays human

For all the hand-wringing, some parts of texting remain very hard to fake.

In-jokes are the clearest example. One word about a holiday three years ago means nothing outside the group. A chatbot was not on the holiday. Shared context works the same way. Your best friend knows that “fine.” with a full stop means you are not fine.

Timing matters too. A reply at 2am says something a reply at noon does not. So does a message the morning after a big day, just asking how it went.

And then there is the single emoji. Knowing when a crying-laughing face is enough. Knowing when a heart beats a paragraph. Knowing when the right response is silence for a while. That judgement comes from knowing a person, not from predicting the next likely word.

AI will keep getting better at writing, and keep changing how we text. But the messages people treasure are rarely the most polished ones. They are the ones only one person could have sent. If you want to keep up with the words we invent along the way, the full dictionary is a good place to start.

Questions people ask

Can you tell if a text was written by AI?

Not reliably. Tells like polished structure, the em dash or “I hope this message finds you well” were common in human writing long before chatbots, so usually the only honest way to find out is to ask.

Why do people type in lowercase on purpose?

Lowercase, typos and abbreviations make a message feel casual and personal. Now that polished writing can look machine-made, a little mess signals that a real person typed it.

Is it wrong to use AI to write a personal message?

Not necessarily. Using a chatbot for a starting point and rewriting it in your own words is like asking a friend for help, but sending something you do not mean tends to feel hollow.

What does clanker mean?

Clanker is a dismissive slang word for robots, chatbots and AI. It comes from Star Wars, where it was an insult for battle droids, and is mostly used jokingly.

What is AI slop?

AI slop is low-quality content churned out in bulk by AI, such as strange images, generic articles and uncanny videos. The word suggests something cheap and not worth your attention.

AI slop

2026-09-12T00:00:00.000Z

What it means

AI slop is content mass-produced with AI, with little effort and no care, mainly to grab attention, clicks or money.

It covers text, images, video and audio, and you’ve almost certainly scrolled past it. Think of surreal pictures on social feeds, like a religious figure made of shrimp. Think of padded articles written by a large language model that take a thousand words to answer a yes-or-no question, often with made-up facts nobody caught. Add product listings with fake photos and near-identical reviews. Nobody who cared looked at any of it before it went out.

“Slop” is doing a lot of work here. It’s an old word for the scraps and swill fed to pigs, the kind of food nobody with a choice would eat. Calling AI content slop says it’s been poured out for an audience assumed to swallow anything.

It’s worth being fair about what slop isn’t. Something made with AI help isn’t automatically slop. An article where AI drafted a section and a person rewrote and checked it doesn’t count, and neither does an image someone spent hours refining. Slop is about low effort and zero care, not about which software was involved.

Where it came from

“Slop” has long meant pig feed, and by extension anything cheap and worthless. The usual account is that people on forums and in AI art circles started calling bad machine output slop around 2022, as image generators opened to the public.

It went mainstream in 2024. That year, tech writers argued that slop should become the standard word for unwanted AI content, the same way spam became the word for unwanted email. One widely cited version of that argument came from developer and blogger Simon Willison in May 2024, who was clear that he hadn’t invented the word. News outlets picked up the comparison, and within months “AI slop” was everyday vocabulary.

The spam comparison explains why it stuck. Spam never meant “email sent by software”. It meant email nobody asked for. The problem isn’t that a machine was involved. It’s being buried in stuff nobody wanted, made by people who didn’t care. That’s a textbook example of how slang spreads: a short, ugly word that fits a feeling lots of people already had. It sits alongside vibe coding and clanker as slang the internet invented about AI.

AI-assisted content	AI slop
A person decides what it should say	Nobody decides much at all
Facts get checked before it goes out	Errors go out unchecked
Made for a particular audience	Made for an algorithm
Published at a human pace	Published by the hundred
Someone puts their name to it	Often from a throwaway account

How people actually use it

“My feed is just AI slop now.” A complaint after scrolling past several fake images in a row.
“Skipped the first five results, all SEO slop.” Search frustration. SEO slop is written to rank, not to help.
“Those reviews are slop, they all say the same thing in the same weird way.” Spotting fake reviews on a shopping site.
“I used AI to tidy the layout, but it’s not slop, I wrote every word.” A creator heading off the accusation before it arrives.
“Mum’s shared another slop picture on Facebook.” A weary, mostly affectionate complaint about relatives being taken in.

In a sentence

Rhi: look at this cabin in the woods, I need to go
Kez: the chimney is coming out of the lake
Rhi: oh no
Kez: it’s slop, mate

Commenter: AI slop, unfollowed
Artist: I painted this by hand, here’s the timelapse
Commenter: ok my bad

Common mistakes

Calling anything made with AI slop. The word is about low effort and no care. A carefully made piece that used AI somewhere along the way doesn’t qualify.
Using it for every bad thing online. People made plenty of rubbish before AI. Slop points at mass-produced machine output.
Assuming you can always spot it. Obvious tells like extra fingers or garbled text are fading as the tools improve. Signs of zero effort are often a better clue than visual glitches.
Accusing real artists without checking. Human-made work gets called slop surprisingly often, and a false accusation can cost someone work.

Spam: unwanted messages sent in bulk, the model slop was named after.
Content farm: a site that pumps out cheap articles to catch search traffic.
Engagement bait: posts built to provoke likes, comments and shares rather than to inform.
Dead internet theory: the half-joking idea that much of the web is now bots talking to bots.
Clanker: a mocking slang word for a robot or AI.

Questions people ask

Why is it called slop?

Slop is an old word for pig feed and watery, unappetising food. Calling AI content slop says it’s cheap filler poured out for an audience assumed to accept anything.

Is all AI-generated content slop?

No. AI slop means low effort and no care. Work where a person used AI but still made the decisions, checked the facts and took responsibility for the result isn’t slop.

Why can’t platforms just remove AI slop?

AI slop is cheap to make and arrives in huge volumes, far more than moderators can review. Detection tools are unreliable, much of it breaks no specific rule, and some of it earns engagement, which gives platforms mixed incentives.

When did people start saying AI slop?

People on forums were calling AI output slop from around 2022. The term became mainstream in 2024, when tech writers pushed it as the name for unwanted AI content, like spam for unwanted email.

How can you tell if something is AI slop?

Look for signs of zero effort: generic wording, padding, confident errors, no named author and a flood of near-identical posts from one account. In images, odd hands, melted lettering and dreamlike backgrounds are common clues.

The short version

AI slop is mass-produced AI content made with no effort and no care, named after the swill fed to pigs. It became the internet’s word for unwanted AI output in 2024, much as spam did for email. Using AI doesn’t make something slop, but not caring does.

Vibe coding

2026-09-12T00:00:00.000Z

What it means

Vibe coding is building software by describing what you want to an AI and accepting the code it writes, usually without reading it properly.

A vibe coder works more like a client than a programmer. They ask for a feature, run whatever comes back, and paste any error message into the chat for a fix. Plenty of professional developers lean on a large language model all day and still check every line. Vibe coders stop checking and judge the work only by whether it seems to do the job.

People split sharply over it. Fans love it for prototypes, personal tools and weekend projects, because an idea can become a working app in an afternoon. Experienced engineers tend to worry about what happens next. Code nobody has read can hide security holes, expose passwords or grow into a tangle nobody can maintain. An AI can also hallucinate a function or package that doesn’t exist, and a vibe coder may not notice until something breaks.

The phrase works as a verb too. You can vibe code a budgeting app, and the person who does it is a vibe coder. Depending on who’s talking, that’s a proud badge or a gentle put-down.

Where it came from

Unusually for slang, the coiner is known. AI researcher Andrej Karpathy used the phrase in a post on X in February 2025. He described a relaxed way of building small projects where he gave in to the vibes, accepted the AI’s changes without reading them and fed any errors straight back in. He presented it as fine for throwaway weekend projects, not as a way to build serious software.

That caveat got lost quickly. The phrase arrived just as AI coding tools had become good enough for non-programmers to build real apps, and those people needed a name for what they were doing. Within weeks it was all over developer forums, YouTube tutorials and tech news. Newer tools that act as AI agents, editing files and running commands on their own, made the habit even easier. By the end of 2025, vibe coding was turning up in dictionary word-of-the-year coverage.

It’s a neat case study in how slang spreads: a casual remark from someone with a big audience, landing in a gap in the language. The word “vibe” made a technical habit sound relaxed, a bit reckless and slightly funny.

Vibe coding	Careful AI-assisted coding
You describe what you want and accept the result	You describe what you want and review the result
Errors get pasted straight back to the AI	Errors get read and understood
“It seems to work” counts as done	Done means tested and understood
Suits prototypes and personal tools	Needed for anything touching money or personal data

How people actually use it

“I vibe coded a little app to split our holiday costs.” Proud use from someone who can’t program.
“This whole codebase is vibe coded and nobody knows how it works.” A complaint. The software runs, but nobody can safely change it.
“Classic vibe coder, put the API key in a public repo.” Mockery of a security slip a careful developer would catch.
“I’m not vibe coding, I read every change.” A defence from someone who uses AI but wants credit for checking its work.
“Spent Sunday vibe coding and now I’ve got a weather dashboard nobody asked for.” Hobby use, close to what the phrase first described.

In a sentence

Mia: wait, you built that? you can’t code
Jordan: I can’t. I vibe coded it
Mia: so the AI built it
Jordan: the AI and my vibes, yes

Lead dev: who wrote the payment code?
Sam: technically nobody
Lead dev: please tell me it wasn’t vibe coded
Sam: it was a very confident vibe

Common mistakes

Using it for all AI-assisted coding. Vibe coding specifically means not reading or understanding the code. A developer who uses AI but reviews everything is doing something else.
Assuming vibe-coded software is always broken. The code can be perfectly fine. The problem is that nobody has checked, so nobody knows.
Treating “vibe coder” as an insult every time. Plenty of founders and hobbyists use it about themselves with pride. Tone and context decide it.
Thinking it means you can skip clear instructions. Good results still depend on describing what you want well, which is the heart of prompt engineering. Vibe coding is loose about the code, not about the request.

Prompt engineering: writing instructions for an AI carefully so it gives better results.
AI agent: an AI tool that carries out multi-step tasks with little supervision.
Hallucination: when an AI states something false with total confidence.
No-code: building apps with drag-and-drop tools instead of code.
Technical debt: the future cost of shortcuts taken in code today.

Questions people ask

Who coined the term vibe coding?

AI researcher Andrej Karpathy coined the term in a post on X in February 2025. He used it for building small projects by talking to an AI and accepting its changes without reading them.

Is vibe coding bad?

Vibe coding is fine for prototypes and personal tools, where a bug costs little. For software that handles payments, logins or personal data, skipping code review is a real security risk.

Can you vibe code without knowing how to program?

Yes, and that’s much of the appeal. People with no programming background build working apps this way, though they often get stuck when the AI can’t fix its own bug.

Why do engineers worry about vibe coding?

Code that nobody understands is hard to fix, extend or secure. Engineers worry about hidden bugs, leaked passwords and apps that fall apart under real use.

Is “vibe coder” an insult?

“Vibe coder” can be either a boast or a put-down. Hobbyists and founders often use it proudly, while experienced developers sometimes use it to suggest someone doesn’t understand their own software.

The short version

Vibe coding is building software by describing it to an AI and trusting whatever comes back. It’s brilliant for quick experiments and risky for anything serious. The term started with one post in early 2025 and spread because it named something lots of people were already doing.

Clanker

2026-09-11T00:00:00.000Z

What it means

Clanker is a mocking slang word for a robot or an AI, used to show contempt, frustration or just to get a laugh.

It’s an insult built on a sound. Robots are imagined as metal things that clank when they move, so calling one a clanker shrinks it down to a noisy tin can. It now covers far more than walking machines, including chatbots built on large language models, customer service bots, delivery robots on the pavement, self-driving cars and the AI-generated videos filling their feeds.

Most of the time it’s a joke with a point. Someone who calls a chatbot a clanker after it invents a fact isn’t having a crisis. They’re venting about confident rubbish, about a bot glazing them with praise they didn’t earn, or about the feeling that AI is being pushed into every corner of life. Clanker gives that irritation a short, satisfying outlet.

It also works as a badge. Saying clanker signals that you’re sceptical of the AI boom, or at least tired of hearing about it. The target is usually the technology or the companies selling it, not a person.

Where it came from

The usual account is that clanker comes from Star Wars. Clone troopers use it for the battle droids they fight, and it’s most closely tied to the animated series The Clone Wars, which started in 2008. In that setting it’s soldiers’ slang, a dismissive name for an enemy that comes off a production line. For years it stayed a fan word.

In mid-2025 it broke out. Frustration with AI had been building for a while, with chatbots replacing customer service staff, AI slop crowding feeds and worries about jobs. People wanted a single insult for all of it, and clanker was ready-made. It was short, funny, recognisably sci-fi and aimed only at machines, so it felt harmless. TikTok, X and Reddit carried it from niche to everywhere within months. It’s a clear example of how slang spreads: an old fandom word suddenly perfect for a new mood.

Not all of it stayed light. A wave of comedy skits imagined a future where robots faced prejudice, and some drew criticism for copying scenes of real racism closely, with clanker as a thin stand-in for real slurs. Most everyday use has nothing to do with that. It’s simply people mocking AI.

In Star Wars	Online from 2025
Said by clone troopers	Said by anyone fed up with AI
Aimed at battle droids	Aimed at chatbots, robots and AI content
Soldiers’ contempt for the enemy	Mockery, often tongue in cheek
A fan word few outsiders knew	A mainstream anti-AI insult

How people actually use it

“The clanker told me there are two r’s in strawberry.” Mocking a chatbot for fumbling something a child could do.
“Let me speak to a human, not a clanker.” Frustration with automated customer service.
“One of those delivery clankers is stuck on the kerb again.” Aimed at a physical robot out on the street.
“This video is clanker-made, look at the hands.” Calling out AI-generated content in the comments.
“That text was pure clanker behaviour.” Teasing a friend whose message sounds stiff and robotic.
“I still say please to the clanker, just in case.” A joke about staying polite to AI in case it remembers.

In a sentence

Leah: the chatbot has apologised four times and still hasn’t fixed my order
Nat: clankers love an apology
Leah: I just want a human

Ben: the robot vacuum keeps chasing the dog
Ife: rogue clanker
Ben: it’s been locked in the cupboard

Marcus: your reply was very clanker
Jo: I was being professional
Marcus: you told your brother you hope this message finds him well

Common mistakes

Assuming it was invented in 2025. The word is much older. Star Wars fans had it for years before it became a general anti-AI insult.
Thinking it’s only for physical robots. In current use it covers chatbots, AI images, automated voices and anything else that feels machine-made.
Using it on real people as a serious insult. Aimed at a friend’s stiff message, it’s a joke. Aimed at someone in earnest, especially in a way that echoes real slurs, it lands badly.
Reading every use as deep hostility. Most people who say clanker are joking or mildly annoyed. Some use it fondly about AI tools they rely on.

Toaster: the sneering nickname for the robots in Battlestar Galactica, now used loosely for any machine.
Luddite: someone who resists new technology, named after 19th-century English workers who smashed textile machinery.
Butlerian Jihad: from Dune, a war against thinking machines, now joke shorthand for turning against AI.
AI slop: mass-produced, low-effort AI content.
Glazing: praising something far too much, also used when a chatbot flatters you.

Questions people ask

Where does the word clanker come from?

Clanker comes from Star Wars, where clone troopers use it as a put-down for battle droids. It’s most associated with the animated series The Clone Wars.

Why did clanker suddenly become popular?

Clanker surged in mid-2025 as frustration with AI grew. People wanted a quick, funny insult for chatbots, robots and AI content, and the old Star Wars word fitted perfectly.

Is clanker a slur?

Clanker isn’t a slur in the usual sense, because it targets machines rather than people. Some comedy skits drew criticism for using it as a thin stand-in for real slurs, but most everyday use is simply mockery of AI.

Can you call a person a clanker?

People do call each other clankers, usually as a joke about sounding robotic, such as sending a stiff, formal text. Between friends it’s a tease rather than a serious insult.

What’s the difference between clanker and AI slop?

Clanker is an insult for the machine itself, such as a robot or chatbot. AI slop is a name for the low-effort content that AI tools churn out.

The short version

Clanker is the internet’s favourite insult for robots and AI, borrowed from the clone troopers of Star Wars. It took off in mid-2025 as a way to vent about chatbots, robots and AI content. Most of the time it’s just mockery with a sci-fi flavour.

Hallucination

2026-09-11T00:00:00.000Z

What it means

A hallucination is when an AI system presents something false or invented as if it were plain fact.

The classic examples are small and specific. A chatbot invents a book and credits it to a real author. It quotes a study that was never published. It supplies a statistic or a web link that leads nowhere. You’d say it hallucinated the book, the study or the link. It sounds just as calm as when it’s right, which is what makes it awkward.

The cause lies in how large language models work. They’re built to produce the text most likely to come next. They have no separate fact checker. When the model has seen plenty about a topic, likely text and true text usually line up. When it hasn’t, it still produces something likely, and likely isn’t the same as true.

Think of a student who hasn’t read the set book but has to write the essay anyway. They know how essays about novels sound, so they write a fluent one, with quotes that feel right. The quotes aren’t in the book. The student isn’t quite lying, just filling a gap with something plausible and hoping nobody checks.

Where it came from

The everyday word means seeing or hearing things that aren’t there. An earlier technical use came from computer vision research, where “hallucinating” detail meant filling in plausible features, such as sharpening a blurry photo of a face. There, it was often the goal rather than a fault.

By the late 2010s, researchers working on machine translation and automatic summaries were using it critically. A system hallucinated when its output included content the source didn’t support, like a summary quoting a figure the original article never gave.

The usual account is that the word reached everyone else in early 2023, once ChatGPT had millions of users and screenshots of confident nonsense were everywhere. News stories followed, including reports of lawyers who filed court documents citing cases that did not exist. Cambridge Dictionary chose “hallucinate” as its Word of the Year for 2023 because of this AI sense.

Not everyone likes the term. Some researchers argue it makes software sound too human, since hallucinating implies a mind having experiences. Alternatives such as “confabulation” have been suggested, but “hallucination” has stuck.

Here’s how it compares with nearby words.

Word	What’s happening	Intent to deceive?
Hallucination	An AI produces false content that sounds plausible	No
Lie	Someone says something they know is false	Yes
Mistake	An ordinary error, often easy to spot	No
Confabulation	Filling a memory gap with invented detail, unaware it’s invented	No
Misinformation	False information spreading, whatever its source	Not necessarily

How people actually use it

“It hallucinated half the references.” Work chat, after someone checked the citations.
“Is this real or did it hallucinate it?” Now a reflex with any surprising AI answer.
“Grounded in your documents to reduce hallucinations.” Product page wording, usually meaning the tool looks things up first, a method called RAG.
“Answers may contain hallucinations. Verify before relying on them.” A typical disclaimer under an AI tool.
“Chatbot invented legal cases, court hears.” A headline format that became familiar from 2023.
“Ben just hallucinated a deadline that doesn’t exist.” A joke aimed at a human colleague.

In a sentence

Nadia: the summary says the report recommends a four-day week
Chris: it doesn’t. I wrote it
Nadia: so the AI hallucinated that
Chris: with great confidence, yes

Tom: it gave me a link to the paper
Ade: does the link work
Tom: no
Ade: there’s your hallucination

Ellie: can I use it for my essay
Raj: for ideas, sure. for quotes and dates, check every single one

Common misconceptions

It isn’t a rare glitch. Hallucination comes from the basic way these models generate text. Newer models do it less, but none is free of it.
The AI isn’t lying to you. Lying needs an intention to deceive, and the model has none. It produces plausible text, and sometimes plausible is wrong.
A citation proves nothing on its own. Models can invent sources, page numbers and links as easily as facts. Check that the source exists and says what’s claimed.
“Don’t make things up” isn’t a fix. Careful prompt engineering helps a little, and so do tools that let the model search or read your files. None of them removes the risk.

Confabulation: a psychology term for filling memory gaps with invented detail, proposed as a more accurate label.
Grounding: tying an AI’s answers to specific, trusted sources.
RAG: retrieval-augmented generation, where a model looks up documents before answering.
Sycophancy: when an AI tells you what it thinks you want to hear, even agreeing with false claims.
AI slop: low-effort AI content churned out in bulk, often unchecked and full of errors.

Questions people ask

Why do AI chatbots hallucinate?

AI chatbots hallucinate because they generate plausible-sounding text rather than checking facts against a reliable source. When they lack solid information, they still produce an answer, and it can be wrong.

Can you stop an AI from hallucinating?

AI hallucinations can be reduced but not eliminated. Giving the model trusted documents to work from, asking for sources and checking key facts yourself all help.

Is an AI hallucination the same as lying?

An AI hallucination isn’t a lie in the usual sense. Lying means knowingly saying something false, and an AI model has no intention to deceive.

Do newer AI models still hallucinate?

Newer AI models hallucinate less often than early chatbots, especially when they can search the web or read documents you supply. They still do it, particularly with obscure topics, exact numbers and citations.

Why was “hallucinate” Cambridge Dictionary’s Word of the Year?

Cambridge Dictionary picked “hallucinate” as its Word of the Year for 2023 because of its newer AI meaning: a system producing false information.

The short version

A hallucination is when an AI confidently gives you something false, from an invented quote to a source that doesn’t exist. It happens because language models predict plausible text rather than checking facts. Treat AI answers as a first draft and verify anything that matters.

LLM

2026-09-11T00:00:00.000Z

What it means

An LLM, or large language model, is a computer program trained on an enormous amount of text so that it can read and write language.

It’s the engine inside chatbots such as ChatGPT, Claude and Gemini. You type a message, and the model builds a reply one small piece at a time. Each piece is a prediction of what should come next. Those pieces are called tokens, and they’re roughly word-sized.

The closest everyday comparison is the predictive text on your phone. Your keyboard suggests the next word based on what people usually type. An LLM does a similar job, but it has learned from a vast mix of books, websites and code, so it can keep going for pages and follow instructions. At scale, that’s enough to draft emails, summarise reports and write code.

“Large” refers to size in two senses. The model has a huge number of internal settings, often billions, called parameters. And the text it learns from is vast. Nobody writes the rules by hand. The model works them out from examples, which is also why its mistakes can be hard to predict.

Where it came from

“Language model” is an old phrase in computing. For decades it described any system that estimated how likely a word was to follow the words before it. Speech recognition and machine translation relied on these long before chatbots made them famous.

The big shift came in 2017, when researchers at Google published a new neural network design called the transformer. It handled long stretches of text far better than earlier methods. Models built on it, including OpenAI’s GPT series and Google’s BERT, appeared from 2018. The usual account is that “large language model” settled into research vocabulary over the next few years, as each model was trained on more data than the last.

The term went mainstream after OpenAI released ChatGPT in late 2022. Journalists suddenly needed a name for the category rather than the brand, and “LLM” filled the gap. By 2023 it was turning up in job adverts and LinkedIn posts. It now anchors a whole vocabulary of newer terms, collected in our AI jargon glossary.

Here’s how it compares with nearby terms.

Term	What it refers to
AI	The broad field of getting computers to do tasks that seem to need intelligence
Machine learning	The part of AI where systems learn from data instead of fixed rules
LLM	A machine learning model trained on text to read and produce language
Chatbot	The app or window you actually talk to, which may run on an LLM
GPT	OpenAI’s family of models, short for generative pre-trained transformer

How people actually use it

“Which LLM is this running on?” The speaker wants to know whose model sits under the branding.
“We’re adding LLM features to the dashboard next quarter.” Work chat. Usually means a chat box or automatic summaries.
“Don’t paste client data into a public LLM.” A standard line in company IT policies.
“New LLM tops rivals on coding tests.” A typical headline after a model launch.
“LLMs are just fancy autocomplete.” A sceptic’s view that the hype has outrun the technology.
“I asked three LLMs and got three different answers.” A common complaint, and a fair reminder to check things.

In a sentence

Maya: is ChatGPT an LLM or is that the company
Leo: ChatGPT is the app. the LLM is the model inside it
Maya: and OpenAI is the company
Leo: yes. three different things

Dad: does the LLM know about the match last night
Jess: only if it can search the web. otherwise it only knows what it was trained on

Common misconceptions

An LLM isn’t the same thing as AI. It’s one kind. Spam filters and face recognition are AI too, but not language models.
It doesn’t look things up by default. A plain LLM answers from patterns learned in training. Some products add web search or document lookup, often through a technique called RAG.
Fluent doesn’t mean correct. An LLM can produce smooth, confident text that is simply false. This is called hallucination, and it’s the main reason to check anything important.
It isn’t learning from your chat as you type. The model’s knowledge is fixed when training ends. Some apps store notes about you separately, but the model itself isn’t rewiring mid-conversation.

Token: the small chunk of text an LLM reads and writes, often part of a word.
Parameters: the internal numbers adjusted during training. More usually means a bigger model.
Context window: how much text a model can take into account at once.
Prompt: the instruction or question you give the model.
Foundation model: a large general-purpose model that other products are built on.

Questions people ask

What does LLM stand for?

LLM stands for large language model, a type of AI trained on vast amounts of text so it can process and produce language.

Is ChatGPT an LLM?

ChatGPT is an app that runs on large language models made by OpenAI. Calling ChatGPT itself an LLM is loose but widely understood.

How is an LLM different from a search engine?

A search engine finds existing pages and sends you to them. An LLM writes a new answer from learned patterns, which is often more convenient but less reliable on facts.

Do LLMs actually understand what they say?

Researchers genuinely disagree about whether LLMs understand anything. What’s clear is that they work by predicting text, and they can be impressive at some tasks while oddly unreliable at others.

Why do people say LLMs are bad at maths?

Early LLMs were shaky at arithmetic because they treat numbers as chunks of text rather than calculating. Newer models are much better, and many AI tools hand sums to separate software.

The short version

An LLM is a large language model: software trained on huge amounts of text to predict what comes next, which lets it write, summarise and chat. It’s the technology inside ChatGPT, Claude, Gemini and most AI assistants. It’s useful, but it isn’t a fact database, so check anything that matters.

Prompt engineering

2026-09-10T00:00:00.000Z

What it means

Prompt engineering is the practice of writing and refining the instructions you give an AI model so it produces the result you actually want.

A prompt is whatever you type into an AI tool. Prompt engineering is the craft of shaping it: what to ask, what background to include and what the answer should look like. A large language model only sees the text in front of it. When a request is vague, it fills the gaps with the most average guess.

Think of it like briefing a temp on their first morning. “Sort out the files” gets you something. “Put last year’s invoices in date order, one folder per client, and flag any without a signature” gets you what you needed. The temp is just as capable either way. The difference is the brief.

The word “engineering” does some heavy lifting. For most people, it just means writing clearly. At the technical end, it means testing many versions of a prompt and keeping the one that fails least. Here’s the everyday version.

Weak prompt	Stronger prompt	What changed
“Write an email to my landlord.”	“Write a short, polite email asking my landlord when the broken boiler will be fixed. It’s been ten days.”	Purpose and facts
“Summarise this report.”	“Summarise this report in five bullet points for a manager who hasn’t read it. Deadlines first.”	Audience, length and priorities
“Give me dinner ideas.”	“Five vegetarian dinners I can cook in 30 minutes. No mushrooms.”	Clear limits
“Is this contract OK?”	“List any clauses that limit my right to cancel. Quote each one, explain it simply, and say if you’re unsure.”	A specific job and permission to be unsure

Where it came from

The word “prompt” already meant a cue, as in a writing prompt or a computer’s command prompt. When text-generating AI arrived, the text you fed it became the prompt too.

The usual account is that “prompt engineering” took hold around 2020, as early users of large language models noticed how much wording changed the output. Adding a few worked examples before the real question could turn a poor answer into a good one. Later, simply asking a model to reason step by step was found to help with maths and logic.

Then chatbots went mainstream, and 2023 became the year of the “prompt engineer”. Headlines described a new job that needed no coding. A few well-publicised listings offered salaries in the hundreds of thousands of dollars. Plenty of people were sceptical, and the term picked up an ironic edge it still has.

As models improved, the magic-words side faded. Newer models follow plain instructions well. From around 2025, developers increasingly said “context engineering” instead. The hard part, they argued, isn’t phrasing one request. It’s choosing everything the model sees: the conversation so far, its tools, its standing rules, and any documents it fetches through retrieval-augmented generation.

How people actually use it

“Can someone good at prompt engineering fix the support bot? It keeps apologising for everything.” Work chat. Fixing the instructions behind a product.
“Prompt engineer: the new job that pays six figures and needs no coding.” A 2023-style headline. The hype version.
“I’m a prompt engineer now, which means I say please to the computer.” Social post. The ironic version.
“Our system prompt went through forty rounds of prompt engineering before launch.” Engineering blog. The serious, testing-heavy sense.
“Honestly, prompt engineering is just writing a decent brief.” A common view from people who find the term inflated.
“Nobody says prompt engineering any more. It’s context engineering.” Developer talk from 2025 onwards.

In a sentence

Priya: The bot writes three paragraphs when people ask for our opening hours.
Sam: Tell it to answer in one sentence unless asked for more. That’s half of prompt engineering.

Jon: My manager wants me to do a prompt engineering course.
Ade: Save the money. Say what you want, who it’s for and what good looks like.

Common misconceptions

“There are secret magic words.” Some phrasings help, but there’s no password. With current models, clear goals and relevant detail beat clever tricks.
“You need to know how to code.” Not for everyday use. The core skill is explaining a task well.
“A good prompt guarantees a correct answer.” It improves the odds. Models can still make things up, which is why hallucination needs its own safeguards.

System prompt: the standing instructions a company gives its AI before you type anything.
Context engineering: choosing everything a model sees, not just the wording of the request.
Few-shot prompting: showing the model a few examples of what you want first.
Chain of thought: getting a model to work through a problem step by step before answering.
Prompt injection: hiding instructions in text an AI reads, to hijack what it does.

Questions people ask

Is prompt engineering still a real job?

Some roles still carry the title, mostly at AI companies. More often, prompt engineering is now a skill listed inside other jobs.

What makes a good prompt?

A good prompt says what you want, who it’s for and what the finished result should look like. An example of the output helps, and so does telling the model it’s fine to say it doesn’t know.

Is prompt engineering the same as vibe coding?

No. Vibe coding means building software by describing what you want and letting AI write the code. Prompt engineering is the wider skill of writing good instructions for any AI task.

What is the difference between prompt engineering and context engineering?

Prompt engineering focuses on how a request is worded. Context engineering covers everything a model receives alongside it, such as documents, earlier messages, tools and standing rules.

Do I need to learn prompt engineering to use ChatGPT?

You don’t need a course to use ChatGPT or similar chatbots well. Everyday prompt engineering mostly means being specific, giving background and following up when the first answer misses.

The short version

Prompt engineering is the skill of wording requests to an AI so the answers come back useful. It was sold as a glamorous career in 2023, then settled into something ordinary: clear writing, good examples and the right context. A good brief still beats a vague one.

RAG

2026-09-10T00:00:00.000Z

What it means

RAG, short for retrieval-augmented generation, is a technique where an AI system first searches a set of documents for relevant information, then writes its answer using what it found.

A large language model learns from a vast amount of text, and then its knowledge is frozen. It doesn’t know your company’s refund policy or a price that changed yesterday. Asked about them, it either admits it doesn’t know or, worse, guesses.

RAG adds a lookup step. Ahead of time, documents are cut into short chunks. Each chunk becomes an embedding, a long list of numbers that captures roughly what the text means, so similar passages get similar numbers. Your question gets the same treatment. A vector search then finds the chunks whose numbers sit closest to it. Those chunks are pasted in next to your question, and the model answers from them.

The easiest comparison is an exam. A plain model sits a closed-book exam, relying on patchy memory. RAG makes it open-book, so the model can find the right page first. But open-book students still turn to the wrong chapter, misread a sentence or pad an answer with guesses. RAG has the same weak spots, which is why it cuts down hallucination without ending it.

Where it came from

The name comes from a 2020 paper by researchers at Facebook AI Research, the lab now part of Meta. It was titled “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks”, and its first author was Patrick Lewis. The paper paired a language model with a searchable index of Wikipedia, and looking passages up first helped on factual questions.

The basic idea wasn’t brand new, since question-answering systems had paired search with language models before. The paper supplied a clear recipe and a name, descriptive rather than catchy.

The usual account is that RAG stayed mostly a research term until 2023. Then businesses wanted chatbots that could answer questions about their own documents. Retraining a model on private data was slow, expensive and soon out of date. RAG offered a cheaper route: leave the model alone and change what it reads. “RAG pipeline” quickly became everyday developer vocabulary.

Today the term covers almost any setup that fetches information for a model before it answers, including the lookups AI agents make as they work. Here’s how it compares with other ways of getting knowledge into an AI.

Approach	Where answers come from	Keeping it current	Can it cite sources?
Plain model	What it learned in training	Wait for a newer model	Not reliably
Fine-tuned model	Training, plus extra lessons on your data	Train it again	Not reliably
RAG	Documents fetched when you ask	Update the documents	Yes, if built to

How people actually use it

“Our assistant uses RAG to answer from your own help centre articles.” Product page. The bot reads your content rather than winging it.
“The RAG pipeline keeps pulling the 2022 pricing doc. Can someone archive it?” Work chat. Used as an adjective, and a reminder that retrieval is only as good as the documents.
“Is this actually RAG, or did they just paste the whole PDF into the prompt?” Developer forum. Separating real retrieval from brute force.
“We tried fine-tuning first, then switched to RAG because our docs change every week.” Engineering blog. The classic trade-off.
“New: answers now come with sources, thanks to RAG.” Product update. The term as a selling point.

In a sentence

Hana: Why did the help bot tell a customer we do free returns? We stopped that in March.
Raj: The RAG index still had the old policy page. It found it and believed it.

Ellie: Couldn’t we just train the model on our wiki?
Tom: It’d be out of date by Friday. With RAG it reads the current version every time.

Priya: Is this chatbot any good?
Chris: Better than most. It’s RAG-based and links each answer to its source, so you can check.

Common misconceptions

“RAG means the answer is correct.” It means the answer is based on retrieved text. If that text is outdated or off-topic, so is the answer, and the model can still misread a good source.
“RAG retrains the model.” It doesn’t touch the model at all. It only changes what the model is shown when you ask.
“A citation proves the claim.” A cited page might not say what the answer claims. When it matters, click through.

Embedding: a list of numbers representing what a piece of text means.
Vector database: a store built to search embeddings quickly by closeness of meaning.
Chunking: splitting long documents into short passages that can be retrieved one at a time.
Fine-tuning: training an existing model further to change how it behaves.
Grounding: tying an AI’s answer to specific sources rather than to its general memory.

Questions people ask

Does RAG stop AI from making things up?

No. RAG reduces made-up answers by giving the model real sources to work from, but the model can still retrieve the wrong passage, misread the right one or fill gaps with guesses.

What is the difference between RAG and fine-tuning?

Fine-tuning changes the model itself through extra training. RAG leaves the model unchanged and hands it relevant documents each time a question is asked, so the information is easy to update.

What is a vector database?

A vector database stores embeddings, the lists of numbers that represent what each chunk of text means. It finds the chunks closest in meaning to a new question quickly, which is why many RAG systems use one.

Do longer context windows make RAG unnecessary?

Longer context windows let a model read more text at once, measured in tokens, but most document collections are still far too large to paste in whole. RAG remains useful for picking out the relevant parts and keeping costs down.

How do you pronounce RAG?

Most people say RAG as a single word, like the cloth, rather than spelling out the letters.

The short version

RAG is how many AI tools look things up before answering: find the relevant passages, add them to the prompt, then reply based on them. It keeps answers current and checkable without retraining the model, and it makes mistakes less likely rather than impossible.

Token

2026-09-10T00:00:00.000Z

What it means

A token is a small chunk of text, such as a short word, part of a longer word or a punctuation mark, that an AI model reads and writes.

Large language models don’t read text letter by letter or word by word. A piece of software called a tokeniser first chops your message into tokens. Common words like “the” or “house” usually become one token each. Rarer words get split, so “tokenisation” might come out as something like “token”, “is” and “ation”. The exact split varies by model, and spaces, punctuation and emoji count too.

Picture a fridge magnet poetry kit. It has magnets for common whole words, plus fragments like “un”, “ing” and “ly” for building everything else. A tokeniser works from a similar fixed set of pieces, and the model writes its reply by adding one piece at a time.

The handy rule for English is that a token is about three quarters of a word. So 100 tokens is roughly 75 words. It’s an average, not a law. Code, numbers, unusual names and most other languages use more tokens for the same meaning.

Where it came from

“Token” is an old word for a sign or a stand-in, as in a bus token or a token of thanks. Computing borrowed it decades ago. A compiler’s first job is to break code into tokens such as names, numbers and symbols. In security, a token is a digital key that proves who you are. Crypto later added a meaning of its own.

Many AI tokenisers use a method called byte-pair encoding, which began as a data compression technique. It repeatedly merges the pairs of characters or chunks that appear together most often in a large body of text. That text is usually dominated by English, so English words tend to become whole tokens. Many other languages, especially those in non-Latin scripts such as Hindi or Thai, get chopped into smaller fragments, so the same sentence can need several times as many tokens.

The pricing sense is newer. The usual account is that it took hold around 2020, when companies began selling access to language models through APIs. Tokens were the obvious unit to charge by, because they’re what the model actually processes. Prices are now usually quoted per million tokens, with separate rates for input (what you send) and output (what the model writes). Output typically costs more.

Limits work the same way. A model’s context window is the most tokens it can handle at once, counting your message, attached files, the conversation so far and its reply. Early chat models managed a few thousand. Many current ones handle hundreds of thousands, and some pass a million.

Here are rough sizes for ordinary English text.

Text	Rough size in tokens
A short, common word	1
A long or unusual word	Often 2 to 4
A 75-word paragraph	About 100
A 750-word article	About 1,000
A 90,000-word novel	About 120,000

How people actually use it

“How many tokens is that prompt?” A developer checking cost, or whether the text will fit. Trimming prompts is a routine part of prompt engineering.
“We hit the token limit halfway through the contract.” The document was too long for the model’s context window.
“Input’s cheap. It’s the output tokens that add up.” Work chat about an AI bill.
“This agent burns through tokens.” A complaint that an AI agent, which works through many steps on its own, runs up a big bill.
“It’s doing about 80 tokens a second.” Speed talk. Tokens per second measures how fast a model writes.

In a sentence

Kai: why did the AI bill triple
Rosa: someone left the agent running all weekend. millions of tokens

Lena: it forgot what I told it at the start
Omar: long chat? you probably went past the context window
Lena: in English please
Omar: too many tokens. it ran out of room

Ines: same prompt in Hindi costs more. is that a bug
Dev: no, Hindi just splits into more tokens

Common misconceptions

A token isn’t a word. It’s often a piece of one. On average, English runs at about four tokens for every three words.
Tokens aren’t only about money. They also set how much a model can read at once and how long it takes to reply.
Models don’t see individual letters. Because they work in chunks, models can stumble on tasks like counting the letters in a word.

Tokeniser: the software that splits text into tokens, spelled “tokenizer” in most documentation.
Context window: the most tokens a model can handle at once, input and output combined.
Input and output tokens: what you send and what the model sends back, usually priced separately.
Tokens per second: a common measure of how fast a model produces text.
Embedding: the list of numbers a model turns each token into before working with it.

Questions people ask

How many words is 1,000 tokens?

In English, 1,000 tokens is roughly 750 words. The exact figure depends on the text, and code or other languages usually fit fewer words into the same number of tokens.

Why do AI companies charge by the token?

AI companies charge by the token because tokens are what their models actually process, so the count tracks the computing work closely. Longer prompts and answers cost more to run.

Why do some languages use more tokens than English?

Tokenisers are mostly built from English-heavy text, so English words often become single tokens. Many other languages get split into smaller pieces, which means more tokens, higher costs and less room in the context window.

Are AI tokens the same as crypto tokens?

AI tokens and crypto tokens are unrelated. An AI token is a chunk of text a model reads or writes, while a crypto token is a digital asset recorded on a blockchain.

The short version

A token is the chunk of text an AI model reads and writes, usually a short word or part of a longer one. In English, a token is about three quarters of a word. Tokens decide what you pay, how much a model can read at once and how fast it replies.

AI agent

2026-09-09T00:00:00.000Z

What it means

An AI agent is an AI system that works towards a goal by choosing steps, using tools and acting on the results, rather than only replying to messages.

A chatbot answers, and you decide what happens next. An agent is handed a task and takes the next steps itself. It might search the web, look something up in company files using RAG, run code or send an email. Then it checks what happened and picks its next move.

Most agents are built around a large language model. The model decides, and the tools act. A loop ties them together: the model chooses an action, software carries it out, the result goes back, and the cycle repeats until the task is done or something breaks.

Think of the difference between asking a friend for directions and asking them to drive you there. Directions are advice. Driving means they make the turns and pick a new route when a road is shut. That’s more useful, but you’re trusting their judgement for the whole trip. How much control you hand over is what people mean by levels of autonomy.

Type	What it does	Who checks the work
Chatbot	Answers questions in conversation	You, before acting on the answer
Assistant with tools	Uses a tool when asked, such as a web search	You, after each request
Supervised agent	Carries out several steps, pausing before risky ones	You, at checkpoints
Autonomous agent	Works through a whole task alone	Often nobody, until it’s finished

Where it came from

“Agent” is an old word in AI research. The usual account is that textbooks from the 1990s onwards framed AI as the study of agents: anything that senses its surroundings and acts on them. By that definition, even a thermostat is a very simple agent.

The modern sense arrived with large language models. Developers began connecting chatbots to tools and letting them run in loops. In 2023, open-source projects such as AutoGPT drew wide attention by letting a model work through its own task lists. They were exciting to watch and unreliable in practice, often going round in circles.

As models got better at using tools, “agent” became the industry’s favourite product word, and from 2024 it was all over launch announcements. The adjective “agentic” came with it. It already existed in psychology and sociology, meaning having agency. Tech borrowed it to mean “working like an agent”.

The word also got stretched. Some products sold as agents are ordinary chatbots or fixed automations with a new label. Critics call this “agent washing”, after greenwashing. Real agents can be useful, especially in software, where coding agents power a lot of vibe coding. The label alone doesn’t tell you much.

How people actually use it

“Meet your AI agent: it books meetings, chases invoices and updates your CRM while you sleep.” Product page. The marketing pitch: a digital employee.
“We’re piloting an agent for first-line support, but refunds still go to a human.” Work chat. A supervised agent with firm limits.
“Is it actually agentic, or is it a chatbot with a button?” Social post. Scepticism about the label.
“Tech firms race to build AI agents that can use your computer for you” News headline. The general-audience framing.
“I asked the coding agent to fix the failing tests and it deleted them.” Developer joke with a real point: agents can meet the letter of a task and miss its intent.

In a sentence

Nina: The vendor says their product is an AI agent.
Omar: Ask what it does without a human clicking a button. If the answer is nothing, it’s a chatbot.

Jess: Did the agent finish the report?
Kai: It drafted it, pulled the sales figures and asked before emailing the client. Exactly right.

Ben: Everything’s “agentic” now.
Lou: My toaster is agentic. It decides when the toast is done.

Common misconceptions

“An AI agent is a robot.” Most agents are just software. They act through apps, browsers and code, not arms and legs.
“Agents have their own goals.” In practice, the goal comes from whoever set the task. The model picks each step to serve that goal.
“More autonomy is always better.” Every unchecked step lets a mistake snowball. An agent can hallucinate a fact early and then act on it repeatedly. Often, an agent that asks first is the better design.

Agentic: the adjective, for AI that plans and acts rather than only responds.
Tool use: a model calling outside functions such as search or email, also called function calling.
Agent loop: the repeating cycle of choosing an action, doing it and reading the result.
Human in the loop: a design where a person approves key steps before they happen.
Multi-agent system: several agents with different roles, passing work between them.

Questions people ask

What is the difference between a chatbot and an AI agent?

A chatbot replies to messages and leaves any action to you. An AI agent is given a goal and takes steps towards it itself, such as searching, editing files or sending messages, then chooses its next move from the results.

What does agentic mean?

Agentic describes AI that acts with some independence, planning steps and using tools to finish a task. An “agentic workflow” is one where AI carries out several steps rather than producing a single answer.

Are AI agents safe to use?

An AI agent is only as safe as the permissions it’s given and the checks around it. An agent that reads your calendar is low risk, but one that can spend money or send emails needs approval steps and limits.

Why does every AI product call itself an agent now?

“Agent” became a strong selling word from 2024, so many companies apply it to products that are really chatbots or simple automations. The useful question is what the product can do on its own, and who checks its work.

The short version

An AI agent is AI that doesn’t stop at answering. It takes a goal, picks steps, uses tools and acts on the results, with more or less supervision. The idea is old and useful, but marketing stretches the label, so ask what a so-called agent can really do by itself.

70 text abbreviations, decoded

2026-09-09T00:00:00.000Z

Most abbreviation lists are sorted A to Z. That works if you already know the letters you are looking for. It fails completely when someone sends you “icl gts, wyd tmrw” and you have to work out what just happened.

So this list is grouped by job instead. Reactions sit with reactions. Time and plans sit together. The ones that carry two different meanings get their own section at the end. Grouping helps because a chat message is mostly context. If you can tell a message is about plans, you only need to scan twelve entries rather than seventy. Every row has a realistic example, and the full dictionary has longer explanations for anything you want to read properly.

Reactions and agreement

These are the smallest units of texting. They tell the other person you read the message and how you feel about it, and most of them have drifted a long way from their original meaning. “lol” almost never signals laughter now. It softens a message so it does not land too hard. The same goes for “lmao”, which mostly means mild amusement rather than anything on the floor. Treat this group as tone, not content.

Short	Means	Used like
lol	Laughing out loud, now a softener	“sorry I forgot lol”
lmao	Laughing hard, usually mild amusement	“lmao that’s so him”
ijbol	I just burst out laughing	“ijbol not the hat”
smh	Shaking my head	“smh he did it again”
ofc	Of course	“ofc I’m coming”
jk	Just kidding	“you’re paying. jk”
bet	Agreed, deal, sounds good	“bet, see you at 8”
w	A win, something good	“that’s a w for us”
l	A loss, something bad	“taking the l on this one”

Honesty and opinion markers

This group flags that an opinion is coming. English speakers have always done this out loud with phrases like “to be fair” or “if I’m honest”, and texting compressed them. The interesting part is that most of these soften rather than sharpen. Someone typing lwk or icl is usually about to admit something small and slightly awkward. Read them as a warning that the real message is in the next few words.

Short	Means	Used like
ngl	Not gonna lie	“ngl I forgot your name”
tbh	To be honest	“tbh I’d rather stay in”
icl	I can’t lie	“icl that was funny”
imo	In my opinion	“imo the first one was better”
imho	In my humble opinion	“imho we’re overthinking it”
fr	For real	“fr, it’s freezing”
istg	I swear to God	“istg if he’s late again”
afaik	As far as I know	“afaik it starts at 7”
lwk	Low-key, quietly, a little	“lwk nervous about tmrw”
idk	I don’t know	“idk, ask her”
idc	I don’t care	“idc where we eat”
tldr	Too long, didn’t read	“tldr: he quit”

Logistics and plans

This is the most useful group to learn and the least interesting. It is pure admin. These abbreviations exist because typing “I am on my way” while walking is annoying. Most of them come from an older era of texting and email, which is why they survive at work as well as in group chats. If you only memorise one section, make it this one, because a misread here means you turn up at the wrong time.

Short	Means	Used like
brb	Be right back	“brb, someone’s at the door”
gtg	Got to go	“gtg, class starting”
otw	On the way	“otw, five mins”
eta	Estimated time of arrival	“what’s your eta?”
ttyl	Talk to you later	“ttyl, have fun”
rn	Right now	“can’t talk rn”
tmrw	Tomorrow	“see you tmrw”
asap	As soon as possible	“call me back asap”
eod	End of day	“I’ll send it by eod”
hmu	Hit me up, contact me	“hmu when you land”
btw	By the way	“btw I moved the meeting”
fyi	For your information	“fyi the shop shuts at 6”

Feelings and flirting

These do two jobs at once. They open a conversation and they signal warmth. “wyd” is the clearest example, because it is rarely a genuine question about your activities. It means someone wants to talk to you. The same applies to a morning “gm” from a person who has never sent one before. If you want the vocabulary that surrounds this stage, rizz covers the charm side and delulu covers what happens when the reading goes too far.

Short	Means	Used like
wyd	What are you doing	“wyd later?”
hbu	How about you	“not much, hbu”
wbu	What about you	“I’m free Friday, wbu”
ily	I love you	“ily, drive safe”
ilysm	I love you so much	“ilysm thank you”
bae	Before anyone else, a partner	“out with bae”
bff	Best friend forever	“she’s my bff”
hbd	Happy birthday	“hbd!! have a good one”
gm	Good morning	“gm, hope today’s easier”
gn	Good night	“gn, talk tmrw”

Compliments and insults

Praise and criticism have both got shorter and blunter. A single letter can now carry a full verdict. Note how many of these are borrowed from performance language, because doing something well is framed as a show. The insults tend to be about being generic rather than being bad, which is a real shift. Calling something “mid” stings more than calling it terrible. Nearby words worth knowing include glazing for over the top praise and ick for sudden, unexplainable turn off.

Short	Means	Used like
ate	Did it perfectly	“she ate that presentation”
slay	Did brilliantly	“new haircut, slay”
mid	Average, forgettable	“the sequel was mid”
sus	Suspicious, off	“that email looks sus”
goat	Greatest of all time	“he’s the goat honestly”
npc	Someone with no original thoughts	“npc behaviour tbh”
ratio	Getting more criticism than support	“you got ratioed”
cooked	In trouble, or finished	“I’m cooked for this exam”

Internet and gaming

Gaming gave texting a lot of its shorthand, and most of it leaked out into normal conversation. You will see people say something is “op” about a coffee machine. Voice chat is where these were built, so they favour speed over clarity. Two of them, “tfw” and “mfw”, started as captions under images and now work as sentence openers. If a friend has vanished mid conversation, “afk” is usually the answer, and if they will not stop typing, that is yap territory.

Short	Means	Used like
afk	Away from keyboard	“afk 5, making tea”
gg	Good game, well played	“gg everyone”
glhf	Good luck, have fun	“glhf on the interview”
op	Overpowered, too strong	“that character is op”
nerf	Weaken something	“they nerfed my main”
irl	In real life	“have you met him irl?”
tfw	That feeling when	“tfw the bus is early”
mfw	My face when	“mfw he says he’s otw”
nsfw	Not safe for work	“nsfw, don’t open at your desk”

The confusing ones

Every abbreviation here has at least two live meanings, and the wrong reading changes the message completely. Context does most of the work. A message about tonight and a message about a phone call will point “ft” in different directions. Terms like hn and mbn also carry a tone that the literal expansion hides, so treat the meaning column here as a starting point rather than a translation.

Short	Means	Used like
ft	FaceTime, or featuring	“wanna ft?”
dm	Direct message	“dm me the link”
pm	Private message, or afternoon	“pm me” / “3pm works”
gts	Go to sleep, or get that s***	“gts, it’s late”
iykyk	If you know, you know	“iykyk 😭”
dw	Don’t worry	“dw about it, sorted”
wtv	Whatever	“wtv, your call”
nvm	Never mind	“nvm, found it”
cc	Copy someone in openly	“cc’ing my manager”
bcc	Copy someone in privately	“bcc me on that reply”
mbn	Must be nice	“holiday again? mbn”
hn	Hell nah, a firm no	“hn, I’m not doing that”

The ones people get wrong

ft gets read as “for the” or “featuring”, which is right on a music track and wrong in a text. In a message from a friend it almost always means FaceTime. Read more on ft if the context is still unclear.
gts looks harmless and often is, because it usually means “go to sleep” at the end of a late night chat. It also has a rude second meaning, which is the one parents tend to find first. The gts page separates the two properly.
pmo is regularly misread as “put me on”, especially when it appears next to “ts”. It normally means annoyed. The current usage sits closer to irritation than anger, and pmo explains how it drifted.
lol is the one adults get wrong most often. Typing “lol” after your own bad news does not mean you find it funny. It means you are trying to make the message easier to receive.
gng is not a warning sign. It is short for gang and it is used the way older generations used “mate” or “bro”. The gng entry has examples of it being used with total affection.
mid is worse than it sounds. It is not neutral. Calling a film, an outfit or a restaurant mid is a real criticism, because being unremarkable is the insult.

How to reply when you have no idea

Answer the part you did understand and let the rest go. “gtg, ttyl” only needs a “no worries, later”. You do not have to decode every letter to reply correctly.
Ask plainly. “what does that mean” is a completely normal message and nobody is offended by it. Guessing wrong in a group chat is far more awkward than asking.
Match the length, not the vocabulary. If you get four words, send four words back. Using slang you are not comfortable with is more obvious than using none.
Check the tone before the meaning. Most short messages are about mood. If someone tells you they are about to crash out, the specific words matter less than the fact they are stressed.

Questions people ask

Are text abbreviations the same as slang?

Not quite. Abbreviations are shortened versions of ordinary phrases, so “brb” and “asap” just save typing. Slang carries attitude and often has no long form at all, which is why a term like lock in needs a proper explanation rather than an expansion.

Which of these are safe to use at work?

The admin ones travel fine. “asap”, “eod”, “fyi”, “cc” and “bcc” have been in office email for decades and nobody blinks. Anything from the reactions or insults sections is best kept for people you actually know.

Why do some abbreviations have two meanings?

They were invented separately by different groups and then collided. “pm” came from email and messaging boards at roughly the same time, and neither side gave way. The reader sorts it out from context, which is why these live in their own section above.

Do abbreviations change quickly?

The logistics ones barely move. “brb” has meant the same thing for about thirty years. The reaction and insult groups turn over much faster, because their whole value is signalling that you are current.

Is it rude to reply with just “k”?

Usually, yes. A single “k” reads as annoyed to most people, where “ok” or “okay” reads as neutral and “kk” reads as friendly. It is a small thing that causes a surprising number of arguments.

How a word gets from one comment to every group chat

2026-09-08T00:00:00.000Z

In 2021, rizz was a small word used by a small number of people on Twitch. It meant something specific: the verbal skill of talking someone into liking you. Two years later it was in newspaper headlines and in the mouths of people who had never watched a stream. Oxford named it word of the year. By 2024 it was mostly a joke about itself.

That journey took about thirty months. Some words do it in six weeks. The interesting part is that the route almost never changes. Slang moves through the same five stages in the same order, and once you know them you can look at any new word and guess roughly how long it has left.

Stage one: the in-group

Every slang term starts somewhere small, inside a group that already shares context. Gaming lobbies and Twitch chat. Drill and grime scenes in London and Chicago. K-pop fandoms. A subreddit with forty thousand members. A school year group.

The single largest source of English internet slang is African American Vernacular English, a rule-governed dialect with its own grammar and a long history of invention. Most of the words in this article came from it. The pattern that follows is just as consistent: terms spread widely, get relabelled as “Gen Z slang,” and the community that made them gets no credit.

The important thing about stage one is precision. Glazing originally described a specific kind of excessive praise, usually of an athlete or a streamer, by someone whose judgement had been compromised by loyalty. It was not a general word for compliments. It had edges.

Most words never leave this stage. Thousands of terms are circulating in small communities right now that nobody outside will ever hear, and that is the normal outcome, not a failure.

Stage two: the export

Then someone builds a bridge.

Usually it is one account with a foot in two worlds: a creator fluent in the source community but with an audience far wider than it. Sometimes it is a single video pushed past its usual audience by the recommendation system. Either way the word lands in front of people who have no idea where it came from.

The comment section does the rest. Someone asks what it means. Three people answer, two of them wrongly. A stitch or a duet repeats it to a fresh audience, and a reaction video explains it badly. Within days the word is being used by people who have only ever seen it used, never heard it used properly.

Meaning starts to drift here, and the mechanism is simple: new users copy the form without the context. They can see the shape of the word and where it sits, so they can produce something that looks right. What they cannot see is the situation that made the original use funny or cutting. So delulu, which began in K-pop fandom as sharp self-mockery about parasocial fantasies, gets flattened into a general word for optimism. The edges come off.

Stage	What is happening	How the word feels
One: the in-group	Coined and used inside a community with shared context	Precise, load-bearing, invisible to outsiders
Two: the export	A bridge account or single clip carries it to a general audience	Novel, slightly confusing, meaning already blurring
Three: saturation	Everywhere at once, stacked and templated	Fun, overused, starting to grate
Four: adult adoption	Brands, teachers, parents and headlines use it	Dead on arrival, faintly embarrassing
Five: irony or fossil	Kept as a knowing joke, or absorbed into ordinary speech	Either winking, or completely unremarkable

Stage three: saturation

At saturation the word stops being information and starts being texture. It is in captions, in replies, in the titles of videos that have nothing to do with it. People use it because it is available, not because it is the right word.

Two things mark the peak.

The first is stacking. Terms start appearing in chains, because each one has become so lightweight that you need several to carry any weight at all. “icl ts pmo” is a stack: an honesty marker, a demonstrative and a complaint, welded into one unit. The same happens with softeners, where “lwk about to crash out” puts a hedge in front of a strong statement so it can be walked back.

The second is templating. The word gets a fixed frame around it. “X is so Y.” “Not the Y.” “Y era.” Once a word only appears inside a template, it has stopped being a word and become a slot filler.

Peak usage is usually the start of the decline. The value of slang comes partly from scarcity: a word that marks you as knowing something only works while there is something to know. At peak, that value has already been spent.

Stage four: the adult adoption

Now the outer ring arrives. A supermarket account uses it in a post about sandwiches. A teacher uses it in a lesson to seem approachable. A newspaper runs it in a headline with quotation marks around it. A parent uses it at dinner, correctly, which is somehow worse than using it wrongly.

People call this adults ruining the word, but that is not what happens. Slang does two jobs. It says a thing, and it says something about who is saying it. The second job dies here. If a word can be used fluently by a brand, a broadcaster and your uncle, it no longer tells anyone anything about the speaker.

Notice that the word usually still means what it meant. It works fine. It just costs nothing to use, so it buys nothing.

The response is rarely a debate. It is a quiet, collective drop. The people who used the word first stopped around late stage three, which is exactly when everyone else was starting.

Stage five: irony or fossil

Words end in one of two places.

Some survive ironically, used with a visible wink by people signalling that they know the word is finished. “Yeet” lives here, and so does “sus,” along with most 2020 slang, kept alive as a joke about itself. Irony is a real afterlife and can last years, but the word is doing a different job: it is a reference, not a description.

Others fossilise into ordinary vocabulary and stop being slang at all. This is the quieter, more successful ending. “Cool” did it. “OK” did it. “Ghosting” did it inside a decade and now appears in advice columns with no explanation. Yap is well down this road, because it fills a genuine gap: English had no compact verb for talking at length to no purpose. Words that name something the language actually lacked tend to survive. Words that were mainly a badge tend not to.

How to tell which stage a word is in

Where you first saw it. Inside a community you belong to means stage one. On a clip from a community you do not follow means stage two already.
Whether people are explaining it. Explainer videos and “what does this mean” comments are the signature of stage two. Nobody explains an in-group word to the in-group.
How many meanings it has. One precise sense means early. Three loose overlapping senses means saturation.
Whether it is stacked or templated. Chains of slang and fixed frames are stage three markers.
Who is using it in your feed. Brands, institutions and headlines are stage four. That is not a comment on them. It is a measure of how far the word has travelled.
Whether it comes with a wink. Deliberate, self-aware use, with a slight pause around it, means stage five.

Why the cycle keeps speeding up

Slang has always worked this way. What changed is the clock, and four things drive it.

Recommendation is the biggest. Older slang spread through connection: you heard it from someone you knew. Recommendation spreads by similarity, so a clip can jump from a small community to millions of people with no link to it. Stage two once took months of person-to-person contact. A single push can now do it overnight.

Short video compresses the exposure. A phrase in a fifteen-second clip can be seen by the same person eight times in an afternoon. Repetition that once took a year of school corridors happens in one sitting, so a word reaches saturation and exhaustion in the same week.

Audience size matters on its own. When the pool is a few billion people, even a word with narrow appeal finds enough users to look ubiquitous.

The last change is the most underrated: language now spreads mainly by text. Written slang can be copied exactly, with no accent and no voice, by someone who does not know how it is pronounced. That removes the friction that used to slow adoption and keep words tied to their communities. It is also why so much current slang is initialism. Speech does not produce many words like icl. Typing does.

None of this is decay. It is the same process running faster on better infrastructure. The vocabulary turns over quickly, and the underlying grammar stays exactly where it was.

What this means if you just want to keep up

Learn the method, not the word list. Any list is out of date on publication. Working out a word from its position, sound and context is a skill that does not expire, which is the point of our method for decoding unfamiliar slang.
Look up the word you actually met. Reading a full dictionary end to end teaches you little. Looking up one word in the sentence where you found it teaches you a lot, because the sentence supplies the context.
Read the stage before you use the word. If a term reached you through a brand or a headline, using it will say something about you that you may not intend.
Do not chase. Understanding slang and speaking it are separate skills, and only the first is generally useful.

Questions people ask

How long does a slang word usually last?

Most terms that escape their original community run the full cycle in one to three years, and plenty never escape at all. A small number stop being slang entirely and stay in the language for decades.

Does a word die because adults use it?

Not exactly. Adult use is the visible sign that a word has finished spreading, and a word that has spread everywhere can no longer signal who is inside a group. It is a symptom of the decline rather than the cause.

Why does the same word mean different things to different people?

Because meaning drifts at stage two, when new users copy the shape of a word without the situation that produced it. The original precise sense usually survives inside the source community while a looser one spreads outside it.

Is internet slang making people worse at English?

There is no evidence for it. Slang is vocabulary, and vocabulary sits on top of grammar rather than replacing it. People who use a lot of it in messages generally switch registers fine.

Can you tell in advance which words will survive?

Not reliably, but there is one decent signal. Words that fill a real gap, naming something English had no compact term for, survive far more often than words whose main function was marking who was in the know.

Rizz

2026-09-05T00:00:00.000Z

What it means

Rizz is the ability to charm someone you’re attracted to, mostly through what you say and how you carry yourself.

It’s a clipping of “charisma”, but it has drifted somewhere narrower. Charisma describes politicians, teachers and stage performers. Rizz is nearly always romantic. If a friend says you have rizz, they mean you can talk to someone you fancy without freezing up or saying something strange.

The word behaves like a stat in a game. You can have no rizz, mid rizz or unreal rizz. That framing is half the appeal. It turns an awkward, personal thing into a number people can joke about. It also makes failure funny rather than humiliating, because a bad attempt is just a low roll.

Rizz also works as a verb. To rizz someone up is to win them over in conversation. The past tense is “rizzed”, and the person doing it is a “rizzler”.

One thing to notice: rizz is almost never said with a straight face. It’s praise delivered with a grin. Plenty of the time it’s applied to something that visibly did not work.

Where it came from

The word comes from “charisma”, with the middle syllable pulled out and the rest thrown away. That kind of clipping isn’t unusual in internet slang, though taking a chunk from the middle of a word rather than the front is rarer, which is part of why rizz sounded odd and new.

The usual account credits the streamer Kai Cenat and his circle in New York, who were reportedly using it on stream around 2021. It spread from Twitch chat to TikTok over the following two years, helped along by “rizz check” clips and by the “unspoken rizz” idea, where someone attracts people without saying anything at all.

By 2023 it had left the internet entirely. Oxford University Press named rizz its word of the year for 2023, which is usually the point where a slang term becomes safe for adults and therefore slightly less fun for teenagers. Nobody claims to have invented the word outright, and the earliest uses are hard to pin down, so treat any precise origin date with caution.

Here’s how it sits against the older words for roughly the same idea.

Word	What it points at
Rizz	Verbal charm aimed at someone you fancy
Charisma	Broad magnetism, works on crowds
Game	Deliberate flirting technique, older and more calculated
Charm	Pleasantness in general, not necessarily romantic
Aura	Presence and vibe, often silent

How people actually use it

“He’s got unspoken rizz.” He attracts people without trying or talking much. Often said about someone quiet.
“I have negative rizz.” Self-deprecating. The speaker thinks they made things worse by opening their mouth.
“She rizzed him up in about four messages.” Used as a verb. She won him over quickly.
“That’s a rizz attempt if I’ve ever seen one.” Calling out an obvious, slightly clumsy flirt.
“W rizz.” Approval. W means win. The opposite is “L rizz”, which is a compliment turned into a burn.
“Rizz check.” A challenge. Show me how you’d approach someone, right now.

Example conversations

Priya: update, I asked her about her tattoo and she showed me all of them

Dan: ok that’s actually W rizz

Priya: I then said “cool” and walked away

Dan: and there it is

Tom: why does everyone talk to Marcus, he says about nine words a night

Ade: unspoken rizz. he doesn’t need the words

Sam: rizz check, someone’s reading a book you like, go

Ella: “what page are you on, no spoilers”

Sam: that’s fine actually. mid but functional

Common mistakes

Using it in a serious setting. Rizz in a job reference or a wedding speech lands as a joke whether you meant it or not. Keep it casual.
Treating it as pure looks. Attractiveness alone is not rizz. Rizz is what you do with the conversation. Someone can be striking and still have famously bad rizz.
Announcing your own rizz sincerely. Saying “I have great rizz” without irony is the fastest way to prove the opposite. Self-reference works best as a complaint.
Assuming it’s an insult. It isn’t. Rizz on its own is positive. Only “L rizz” or “no rizz” turns it around.

Aura: presence and vibe, gained or lost by your behaviour. Often counted in points.
Rizzler: a person with a lot of rizz. Usually said with affection or mockery.
Simp: someone who gives too much attention to a person who isn’t interested.
The ick: a sudden loss of attraction over a small detail.
Delulu: convinced against the evidence that things will work out for you.
Cooked: finished, out of options. Frequently applied to a failed rizz attempt.

Questions people ask

Is rizz still used, or is it dated?

It’s still current, but it aged fast after 2023. Younger users now often say it ironically, and using it too earnestly can read as trying to sound younger than you are.

Does rizz have to be about flirting?

Usually, yes. You’ll occasionally see it stretched to mean charm in general, like rizzing a bouncer into letting you in, but the romantic sense is the default.

What’s the difference between rizz and aura?

Rizz is active and verbal, aura is passive and atmospheric. You use rizz on someone. You simply have aura, and you can lose it by tripping over.

Can you say it about a girl or a woman?

Yes. The word started out mostly in conversations between young men, but it’s used about anyone now.

How do you spell the verb forms?

“Rizz up”, “rizzed up”, “rizzing”. Double z throughout. There’s no agreed spelling for anything more elaborate, so keep it simple.

Is calling someone a rizzler a compliment?

Mostly. It’s admiring, with a small edge of teasing built in. Nobody uses it in complete seriousness.

The short version

Rizz is short for charisma, narrowed down to mean the skill of charming someone you’re attracted to through conversation. It works as a noun and a verb, and it’s nearly always said with a hint of a joke. If someone tells you your rizz is strong, take the compliment and don’t ask them to explain it.

Delulu

2026-09-04T00:00:00.000Z

What it means

Delulu is a softened, joking version of “delusional”, used for someone whose hopes have drifted away from the facts.

The clinical word is heavy. Calling someone delusional suggests something is seriously wrong. Delulu removes that weight by making the word sound silly. The repeated syllable does the work. It signals that this is teasing, not a diagnosis.

Most of the time people say it about themselves. “I’m being delulu” is a way of admitting you know your expectations are unreasonable, while carrying on anyway. That admission is the whole point. You get to keep the fantasy and stay honest about it at the same time.

It’s an adjective first. You are delulu, or you’re in your delulu era. It also works as a noun, as in “the delulu is strong today”, and occasionally as a label for a person, though “she’s a delulu” is less common than “she’s being delulu”.

Aimed at someone else, the tone depends entirely on the friendship. Between close friends it’s affectionate. Said flatly to a stranger, it’s a put-down.

Where it came from

The usual account traces delulu to English-language K-pop fandom around 2014. It described fans who had convinced themselves they’d end up dating an idol, or that a particular idol was secretly in a relationship they had pieced together from photographs. Those fans were sometimes called “delulu fans”, and the shortening stuck inside those communities for years.

It sat there quietly until TikTok picked it up in 2023 and attached it to the phrase “delulu is the solulu”. That’s mock baby-talk for “delusion is the solution”, often extended to “delulu is the solulu to the trululu”, meaning the true problem. The joke reframed the word. It stopped being about fandom and became a piece of ironic self-help, aimed at job applications, crushes, gym goals and long-shot plans.

That second life is what made it general. Reports in 2025 had it entering mainstream dictionaries alongside other internet coinages, which is the usual sign a slang word has settled. The 2014 origin is fandom folklore more than documented history, so treat the date as roughly right rather than exact.

Here’s how the neighbouring words differ.

Word	The attitude behind it
Delulu	Hopeful, knows it’s unrealistic, proceeds anyway
Manifesting	Sincere belief that expectation shapes outcomes
Copium	Comforting yourself after something already went wrong
Wishful thinking	The plain, unfunny version of the same thing
In denial	Refusing to accept something, usually not self-aware

How people actually use it

“Going delulu about this interview, I’ve already picked my desk.” Announcing unearned confidence for fun before an outcome is known.
“Delulu is the solulu.” The catchphrase. Roughly: unrealistic self-belief is the answer, so commit to it.
“He liked one story and she’s fully delulu now.” Teasing a friend for reading far too much into a small signal.
“I’m in my delulu era.” A stretch of time spent deliberately ignoring the odds. “Era” makes it sound like a chosen phase.
“Slight delulu, but do you think he remembers me?” Used as a hedge, softening a question the speaker knows is a stretch.
“That’s not delulu, that’s just lying to yourself.” Marking the line where the joke stops being a joke.

Example conversations

Nia: he replied “haha yeah” so I think it’s going well

Jo: delulu

Nia: the delulu is the solulu, Jo

Ravi: I applied for the senior role. I have eight months of experience

Sana: genuinely delulu behaviour and I respect it

Ravi: worst case they say no and I’m exactly where I started

Kim: she’s rearranged her whole week around a maybe

Alex: that’s past delulu. someone needs to say something

Common mistakes

Using it about actual delusions. If someone has genuinely lost touch with reality, the joke word is the wrong tool. Delulu is for crushes and job applications, not for mental health.
Saying it to strangers. Without the shared warmth, it reads as calling someone stupid. Reserve it for people who’ll hear the affection in it.
Confusing it with manifesting. Manifesting is sincere. Delulu knows it’s nonsense. Swapping them misses the irony that makes the word work.
Writing it as “delulu” in formal contexts. It has no place in a report or an email to a client, even a friendly one. The whole word is built to sound unserious.

Solulu: the same baby-talk treatment applied to “solution”. Rarely used outside the catchphrase.
Copium: an imagined drug you take to feel better about a loss you can’t change.
Manifesting: the sincere belief that focusing on an outcome helps bring it about.
Main character energy: behaving as though the story revolves around you.
Yapping: talking a lot with little substance, often what delulu people do about their crush.
Rizz: charm aimed at someone you fancy. Frequently the thing the delulu person believes they have.

Questions people ask

Is being called delulu an insult?

It depends on who says it. From a friend it’s teasing and often approving. From someone you barely know it’s dismissive.

Why the repeated ending?

The doubled syllable makes the word sound childish, which strips out the seriousness of “delusional”. Slang does this often, and it’s the same instinct behind “solulu” and “trululu”.

Is delulu always about romance?

No. It started in fandom and it’s still common for crushes, but people apply it to careers, sports predictions, house prices and any plan with poor odds.

Is it healthy to be delulu on purpose?

As a confidence trick before something scary, it’s mostly harmless and can help. It becomes a problem when it replaces actual preparation or stops you accepting a clear answer.

Do people still say it?

Yes, though it peaked around 2023 and 2024. It’s now settled into ordinary casual speech rather than being a fresh phrase.

What’s the opposite?

There isn’t a fixed one. People reach for “grounded”, “realistic” or the blunter “reality check” when they want to mark the contrast.

The short version

Delulu is “delusional” made small and funny, used for hopes that have quietly stopped matching the evidence. It’s usually aimed at yourself, and the joke is that you know exactly what you’re doing. Used kindly it’s a way of cheering someone on. Used coldly it’s just an insult with a sillier spelling.

LWK

2026-09-03T00:00:00.000Z

Someone in the group chat types “lwk kinda miss him.” No capital letters, no explanation, three letters doing a lot of work. If you had to stop and wonder whether that was a typo, you are in good company.

“lwk” is one of the most searched pieces of text slang right now, mostly because it looks like it should be an acronym for something and is not. This guide gives you the meaning up front, then covers where it came from, how people use it on different apps and how to answer without sounding like you had to look it up.

What it means

Put simply, lwk is a compressed spelling of low-key. People use it to soften a statement, admit something quietly or say they feel a certain way without making it a big deal.

Think of it as the texting version of “kind of,” “a little” or “honestly, but don’t make a thing of it.” So when a friend says:

“I’m lwk stressed about tomorrow.”

They are stressed, but they are downplaying it. They are not shouting about it and they do not want a dramatic reaction. They just want it noted.

“Low-key” has been a normal English adjective for decades, describing something relaxed or understated. Texting turned it into a filler adverb, and then people trimmed it to three letters to type faster.

At a glance


Full form	Low-key
Job in a sentence	Softens whatever follows it
Closest plain English	“kind of”, “a little”, “quietly”
Opposite	High-key
Politeness	Neutral, takes its tone from the words around it
Safe at work?	No, keep it to personal messages

Other Meanings of “lwk”

“Low-key” covers the vast majority of uses, but a few other expansions show up depending on who is typing.

Meaning	How common	Example
Low-key	Very common	“lwk want pizza rn”
Last week	Rare	“saw him lwk at the mall”
Light work	Rare, gaming and sports	“that boss was lwk”
Love with kisses	Very rare	Sign-off in older text messages

If it appears mid-sentence before a feeling or an opinion, it means low-key. If it appears after a time reference, it might mean last week. Context settles it almost every time.

How People Use “lwk” in Real Conversations

The word works as a softener. Drop it in front of almost any statement and the statement becomes gentler, more casual and easier to walk back.

Admitting a feeling: “lwk nervous for the results”
Sharing an opinion: “the new album is lwk better than the last one”
Flirting carefully: “lwk been thinking about you”
Reacting to a meme: “lwk me every Monday”
Confessing something small: “lwk forgot we had plans”

Where You Will See It Most

TikTok comments, where “lwk” often opens a hot take.
Snapchat, because typing fast matters more than spelling.
Group chats, where it lets people say something real without a spotlight.
X and Instagram replies, usually in the “lwk true” or “lwk fire” pattern.

Is “lwk” Rude, Flirty or Offensive?

No. It is neutral. The tone comes entirely from the words around it. “lwk hate this” is negative, “lwk obsessed with you” is affectionate, and “lwk tired” is just tired. The abbreviation itself carries no attitude.

It can read as flirty when the rest of the message is flirty, because “low-key” signals a feeling the sender is only half-admitting. That vulnerability is often the point.

How to Respond When Someone Says “lwk”

Match their energy. If they downplayed something, do not blow it up.

They say “lwk nervous about the interview” → “you’ll be fine, lwk jealous of your prep though”
They say “lwk miss you” → “same tbh” or “lwk miss you too”
They say “that show is lwk mid” → “lwk agree” or “nah it’s high-key great”

The opposite of low-key is high-key (sometimes typed “hk”), meaning openly or very. If someone is being low-key about something you are enthusiastic about, “high-key” is the natural counter.

Common Mistakes

Reading it as an acronym. It is not L.W.K. anything. It is a compressed spelling.
Using it in formal writing. Keep it out of emails, essays and anything a manager reads.
Over-using it. One “lwk” softens a message. Three in a row make you sound unsure of everything.

Similar Slang You Should Know

hk / high-key: openly, very much, the opposite of lwk.
ngl: not gonna lie, another honesty softener.
tbh: to be honest, often used right next to lwk.
fr: for real, used to confirm someone else’s low-key statement.
icl: I can’t lie, the UK-flavoured cousin of ngl.

Questions people ask

Someone used lwk twice in one message, does that mean anything?

Not really. Doubling it is just emphasis, the same way people repeat “so so tired”. If anything it signals they are a little more invested than one lwk would suggest.

Can lwk go at the end of a sentence?

Yes, though it is less common. “I miss him lwk” works and lands slightly softer than putting it up front.

Why do people write it in lowercase?

Lowercase is part of the tone. Capitalising it would make the word louder, which defeats the point of a term that exists to turn the volume down.

Is there a version for the opposite feeling?

High-key, sometimes typed hk. If lwk means you are admitting something quietly, high-key means you are announcing it.

Does it work in speech, or only in text?

People say “low-key” out loud constantly. Saying the letters aloud is rarer and usually done as a joke about texting.

Will my message read badly if I use it wrongly?

Almost never. Worst case someone reads a slightly softer version of what you meant, which is a small misfire compared with most slang.

The short version

Once you have seen it in a few messages, lwk stops being cryptic. It is a small word with one job, which is to take the edge off whatever follows. Keep it to casual chats, leave it out of anything formal, and you will read it correctly every time.

Yap

2026-09-02T00:00:00.000Z

What it means

To yap is to talk at length about something nobody asked about. The word carries a mild accusation. If someone says you are yapping, they mean you have been going for a while and the point has not arrived yet.

The gerund does most of the work. “Yapping” is the activity itself, and it can be affectionate or dismissive depending on delivery. “Stop yapping” is a request for quiet. “She was yapping about her ex for an hour” is a report on a long, unfocused monologue. A “yapper” is a person known for doing it, and the label tends to stick. Someone can be introduced as the yapper of the group and nobody needs it explained.

A “yap session” is a different thing again, and it is usually positive. It describes a long, meandering conversation between friends where the length is the whole point. Two people on a three hour call about nothing in particular have had a yap session, and both of them enjoyed it.

The word is rarely a real insult. It sits closer to teasing than to criticism. Calling a friend a yapper is much softer than calling them boring or annoying, and people often claim the title for themselves. “I’m a yapper, sorry in advance” works as a disclaimer before a long voice note.

Where it came from

Yap is an old English word for the short, sharp bark of a small dog. It is imitative, meaning the word was built to sound like the noise. English has used it that way since at least the seventeenth century, and the transfer to human speech followed soon after. By the nineteenth century, yapping already meant chattering in a way that irritated the listener, and yap was American slang for the mouth itself. None of this part is new.

What changed is the frequency and the tone. The current sense took off online around 2023. The usual account puts its build-up in Twitch chat, since a streamer talking through a quiet stretch of gameplay gives chat something to react to, and yap is short enough to spam repeatedly. From there it moved to TikTok, where the joke was formalised into a format: clips captioned “yapping”, fake counters for how long someone had been going, jokes about a yap tracker, and the word “yappacino” for a drink that makes you talk too much.

The revival kept the old meaning and dropped most of the hostility. Older uses of yapping suggested a nuisance you wanted to escape. The current use suggests entertainment you have chosen. That shift is why the word spread so quickly into group chats and Discord servers, where the accusation is nearly always fond.

How people actually use it

“sorry for the yap”: tacked onto the end of a long message. A pre-emptive apology, and not a very sincere one.
“he was yapping the whole film”: a straight complaint, using the older and more irritated sense of the word.
“get in, yap session”: an invitation to a long, aimless catch-up. Entirely positive.
“chat he’s yapping again”: Twitch phrasing, addressed to an audience about a third person, half mocking and half fond.
“certified yapper” or “D1 yapper”: a title rather than a description. The D1 borrows from American college sports and means top tier.
“less yapping more gaming”: a request to stop the commentary and get on with the activity.

Yap overlaps with several older words, but each one implies something slightly different.

Word	What it suggests	Tone
Yap	Long, unfocused talking with no destination	Teasing
Rant	Long, heated talking that does have a point	Sympathetic or wary
Waffle	Talking around a point to fill space	Impatient
Ramble	Wandering off topic, usually harmlessly	Gentle
Glaze	Praising someone far past the point of comfort	Mocking

Example conversations

Priya: sorry that voice note was four minutes Sam: four minutes of pure yap Priya: i did warn you at the start

Alex: so the manager said no, which is fine, but then he starts explaining the budget thing again, and Jordan: you’re yapping Alex: i am building to something

Kofi: free tonight? need to talk about basically everything Mia: yap session. 8pm. i’ll make tea.

Common mistakes

Treating it as an insult. Most of the time it is a nudge, not a verdict. Replying as though you have been attacked reads as missing the joke.
Using it for content that has a clear point. A long but well argued message is not yapping. Length on its own does not qualify, because the word needs that sense of drift.
Confusing a yapper with the older slang sense of yapper as the mouth, as in “shut your yapper”. People still say that, and it means something different from calling a person a yapper.
Aiming it at a stranger or at someone senior to you. It is a friendship word. In a work message it lands as rude, even with a smiley face attached.

Glazing: praising someone so heavily that it becomes uncomfortable to watch.
Waffle: a British term for talking at length without saying much of anything.
Wall of text: a long, unbroken block of writing that nobody wants to read.
Rant: a sustained complaint. It has a target and a point, which a yap does not.
D1 yapper: the top tier of yapper. Delivered as half compliment, half warning.
Chatty Patty: an older nickname for someone who talks constantly.

Questions people ask

Is being called a yapper a bad thing?

Usually not. It works more like a running joke about someone’s personality than a real complaint, and plenty of people apply it to themselves before anyone else can.

Does yapping mean lying or talking nonsense?

No, and that is a common misreading. Yapping is about quantity and direction, not truth. A yapper can be completely accurate and still be yapping.

What is a yap session?

A long conversation between friends with no agenda, in person or on a call. The word session signals that both people knowingly signed up for it.

Why do streamers and their chats use it so much?

Live streams have long stretches where very little happens and the host talks to fill them. Chat needs a short word for that, and yap fits comfortably in a single message.

Can you yap in writing?

Yes. A twelve paragraph message with no clear request counts. So does a voice note that takes four minutes to reach the actual question.

Is yap a new word?

Not at all. It has described a dog’s bark for centuries and human chatter for nearly as long. What is new is how often people reach for it and how friendly it now sounds.

The short version

Yapping is talking a lot without getting to the point, and a yapper is someone who does it often enough to be known for it. The word itself is old, but the warm, teasing version dates from around 2023 and spread through Twitch and TikTok. Use it with friends, and expect to be called one back.

GNG

2026-09-01T00:00:00.000Z

You see it in a TikTok reply: “nah gng that’s crazy.” Then in a group chat: “we outside gng.” It is not capitalised, nobody explains it, and it clearly is not “going” because the sentence still works without it.

“gng” is one of those abbreviations that spread through comment sections before most people noticed. This guide explains what it means, where it came from, the alternative meanings and how to use it without sounding forced.

What Does “gng” Mean in Text?

In almost every case, “gng” means “gang.” It is not about an actual gang. It is a casual way of addressing a friend, the same way people say “bro,” “dude” or “man.” Dropping the vowel just makes it quicker to type.

So “you good gng?” simply means “you good, bro?”

The word “gang” as a friendly address has been part of hip-hop and street culture for a long time. Doubling it, as in “gang gang,” became a greeting and a sign of loyalty. Texting shortened it, and “gng” became the standard spelling in comments and DMs.

Quick Answer

Question	Answer
What does gng stand for?	Gang
What does it mean?	Bro, friend, my people
Is it rude?	No, it is friendly and casual
Where is it used?	TikTok, Snapchat, Instagram, X, group chats
Who uses it most?	Gen Z, especially in hip-hop and sports spaces
Formal or informal?	Very informal

Other Meanings of “gng”

Meaning	How common	Example
Gang (friend, bro)	Very common	“wyd gng”
Going	Occasional typo-style shorthand	“gng to the store”
Good night, G	Rare	“gng, talk tmrw”
Gang gang (doubled)	Common as “gng gng”	“gng gng, we up”

If “gng” sits at the end of a sentence or right after a name, it means gang. If it comes before a place or an activity, it probably means going.

How People Use “gng”

Greeting: “wassup gng”
Agreement: “facts gng”
Reassurance: “you got this gng”
Loyalty: “gng for life”
Reacting to something wild: “gng what did I just watch”

It is often paired with other slang like “ts” (this), “pmo” (pisses me off) and “icl” (I can’t lie), so a full sentence might look like “icl gng ts is wild.”

Is “gng” Offensive?

No. It is friendly. That said, it is tied to hip-hop culture and can feel out of place if you are clearly using it as a costume. If it is not how you or your friends talk, you do not need to force it.

It is also worth knowing that in some school and workplace contexts the word “gang” can be misread. Keep it in casual chats with people who use it too.

How to Reply to “gng”

Treat it as “bro” and reply normally. You can mirror it if that fits your style:

“wyd gng” → “nothing much gng, you?”
“you good gng?” → “all good, appreciate you”
“facts gng” → “fr”

Similar Slang

twin: a close friend, someone on your wavelength.
bro / bruh: the universal casual address.
fam: family, used for close friends.
gng gng: doubled for emphasis or as a greeting.
dawg: an older cousin of gng with the same meaning.

Frequently Asked Questions

What does “gng” mean on TikTok?

It means “gang,” used as a friendly way of addressing someone, like “bro.”

Does “gng” mean “going”?

Sometimes, when it appears before an activity or place. Most of the time it means gang.

Is “gng” a gang reference?

Not in the literal sense. It is a term of friendship borrowed from hip-hop slang.

Can I say “gng” to anyone?

It is best with friends who already talk that way. It will sound odd in formal or professional settings.

Final Thoughts

“gng” is just “gang,” and “gang” is just “bro.” Once you see it that way, every confusing comment suddenly reads normally. Use it with friends, skip it at work, and you are set.

netvocab.com

APR

What it means

Where it came from

How people actually use it

In a sentence

Common misconceptions

Related terms

Questions people ask

What is the difference between APR and the interest rate?

Does the lowest APR always mean the cheapest borrowing?

Why was I offered a higher APR than the advert showed?

Does 0% APR mean the credit costs nothing?

Is APR worked out the same way everywhere?

The short version

Escrow

What it means

Where it came from

How people actually use it

In a sentence

Common misconceptions

Related terms

Questions people ask

Who pays the escrow fee?

Is escrow the same as a deposit?

How long does money stay in escrow?

Can money in escrow be returned?

Does the United Kingdom use escrow?

The short version

AI jargon, translated

The basics

How models learn

How you use them

Building with AI

When it goes wrong

The culture and the slang

Three terms people mix up

How to keep up without reading every launch post

Questions people ask

What is the difference between AI and ChatGPT?

Why do AI chatbots make things up?

What does “tokens” mean on an AI pricing page?

Is prompt engineering a real job?

Is an AI agent just a chatbot with a new name?

How AI is quietly changing the way we text

The autocomplete era

“Did ChatGPT write this?”

Writing messy on purpose

Ghostwritten feelings

The slang people invented about AI

What stays human

Questions people ask

Can you tell if a text was written by AI?

Why do people type in lowercase on purpose?

Is it wrong to use AI to write a personal message?

What does clanker mean?

What is AI slop?

AI slop

What it means

Where it came from

How people actually use it

In a sentence

Common mistakes

Related terms

Questions people ask

Why is it called slop?

Is all AI-generated content slop?

Why can’t platforms just remove AI slop?

When did people start saying AI slop?

How can you tell if something is AI slop?

The short version

Vibe coding

What it means

Where it came from

How people actually use it

In a sentence

Common mistakes

Related terms

Questions people ask

Who coined the term vibe coding?