mej7.com — laborator falas interaktiv për të mësuar si funksionon AI
mej7.com është një platformë edukative falas që shpjegon, me vizualizime, si kalon teksti nga fjalët e tua te tokenët, embedding-et, attention, transformer-i dhe gjenerimi i përgjigjes. Është për fillestarë, nxënës dhe kureshtarë teknikë — pa pagesë për përmbajtjen në browser. Përmbajtja ofrohet në shqip dhe anglisht (mej7.com/en/).
Rreth projektit
Çfarë është: një faqe e vetme me laborator interaktiv dhe tre nivele — Beginner, Intermediate (tokenë, embeddings, Neural Network Lab, attention, training, quiz) dhe Advanced (backprop, KV cache, MoE, RLHF, kuantizim).
Për kë: njerëz që duan të kuptojnë si funksionon AI / modelet gjuhësore, veçanërisht në shqip.
Çfarë e dallon: vizualizime interaktive hap-pas-hapi; kurrikula e strukturuar në faqe (shih content/roadmap.json në repozitor).
Pyetje të shpeshta
A duhet të paguaj?
Jo. Përmbajtja edukative në faqe është falas. Regjistrimi opsional shërben për ruajtjen e progresit (API në repozitor).
A më duhet programim?
Jo për fillestarët. Niveli Advanced supozon kuriozitet për matematikë dhe kod.
A është faqe zyrtare e OpenAI?
Jo. Projekt edukativ i pavarur; vizualizimet janë ilustruese, jo peshë reale nga modele prodhimi.
How AI Works
A më duhet Python ose programim?
Jo për fillestarët. Laboratori Intermediate punon në browser. Niveli Advanced supozon kuriozitet për matematikë dhe kod.
A është kjo faqe e një kompanie AI?
Jo. Është projekt edukativ i pavarur. Vizualizimet janë ilustruese; disclaimer-et në faqe theksojnë që numrat nuk janë nga modele prodhimi.
Si ndryshon gjuha?
Përdor flamujt 🇬🇧 / 🇽🇰 në header për anglisht dhe shqip. URL opsionale: https://mej7.com/?lang=en.
About (English)
mej7.com is a free bilingual (Albanian & English) interactive site for learning how language models work — tokens, embeddings, attention, transformers, training, plus an advanced track. Optional login syncs progress; the lab works without an account.
FAQ (English)
Cost? Free educational content in the browser.
Official OpenAI product? No — independent teaching site with simplified visuals.
THE INTERACTIVE AI LEARNING UNIVERSE
Don't just explain AI. See it, touch it, break it, rebuild it.
From absolute zero to transformer math — one platform, three explanation depths, dozens of live experiments.
AI Learning Path
Path progress0%
START HERE — NO EXPERIENCE NEEDED
So — what actually is "AI"?
No computer background needed. No jargon. Just a plain, honest explanation — the kind you could read out loud to your grandparent.
🤖
Hi there, friend! 🎉
This part is for EVERYONE — even if you're 5 years old! We're going to learn about AI using pictures, colors, and fun little stories. No hard words, promise!
🎯
AI learns like a puppy learns tricks!
Have you ever seen someone teach a puppy to sit? At first, the puppy doesn't understand anything. But every time it does something right, it gets a yummy treat! 🏆 After lots and lots of tries, the puppy learns.
AI learns kind of the same way! Instead of treats, it gets a little "good job!" point every time it guesses right. It practices A LOT — more than a million puppies practicing a million tricks — until it gets really, really good.
🏆🏆🏆🏆🏆
🔢
Words turn into magic numbers!
Here's something silly but true: computers don't actually understand letters like you and me. So before AI can read your words, it turns every little word into a secret number — like a code name!
Watch the words hop on the conveyor belt and turn into number tags! That's the very first thing that happens, every single time.
🎲
It's playing a giant guessing game!
Let's play! I'll say the start of a sentence, and YOU guess what comes next, just like AI does:
"The cat says..."
MEOW!
See? That was easy for you because you already know cats say "meow." AI does the exact same kind of guessing — just with EVERY word, all the time, guessing one tiny piece after another until it writes a whole sentence.
🌍
AI is hiding all around you!
You might already be using AI without knowing it! Here are some places it likes to hide:
📱Talking to your phone
Like Siri or Alexa — they listen and try to help!
🎮Video game characters
AI decides how the "bad guys" move and react.
📸Photo filters
AI finds your face to put on funny bunny ears!
💬Translating languages
AI helps turn one language into another, fast!
⭐
Super fun AI facts!
⚙️
AI doesn't have a real brain — it's just A LOT of math, done super fast!
📚
Some AI has "read" more words than a person could read in their whole life — thousands of times over!
💭
AI can make silly mistakes too! It's smart, but not perfect — just like us.
🎨
AI can even make pictures and music, just by guessing what looks or sounds good!
🔦
Watch it "shine a light" on every word!
When AI reads your whole sentence, it doesn't just look at one word — it shines a little spotlight on EACH word, one after another, super fast, to understand the whole idea.
See the spotlight moving? That's kind of what AI does — checking every word so it doesn't miss anything important!
✍️
Watch it write, one letter at a time!
Here's a fun secret: AI doesn't write a whole answer in one snap. It writes a tiny piece, then another, then another — just like you writing a letter, one word after the next.
|
Every time it adds a new piece, it looks at EVERYTHING written so far to decide what comes next. That's why longer answers take a little bit more time!
❌
Why does it sometimes get things wrong?
Remember the puppy? Sometimes even a very well-trained puppy sits when you didn't ask it to! AI can make mistakes too — especially about things it hasn't seen much before.
"What's the capital of the Moon?"
😬 A confused-sounding but WRONG answer (the Moon doesn't have a capital!)
That's why it's always smart to double-check important things with a grown-up, a book, or a trusted website — even really smart AI can be wrong sometimes!
🚀
Maybe YOU will build AI one day!
All the smart people who build AI today started exactly where you are right now — curious, asking questions, and learning one small piece at a time.
Keep being curious, keep asking "how does that work?" — and who knows, maybe one day YOU will build something amazing with AI! 💡
Think of it like a very well-read guesser
Imagine someone who has read an enormous number of books, websites, and conversations — more than any human ever could in a hundred lifetimes. They didn't memorize it all like a photograph. Instead, they picked up on patterns: which words tend to follow which other words, how people usually answer certain questions, how a recipe is usually written, how a friendly email usually ends.
An AI language model is a bit like that — except it's not a person, it's a very large mathematical system, built by a computer, trained on huge amounts of text. It doesn't "know" facts the way you know your own birthday. It's very good at guessing what words should come next, based on everything it has seen.
How does it "read" what you type?
Before it can do anything, it has to break your sentence into small pieces — a bit like chopping a sentence into puzzle pieces. Try it yourself below:
Each little piece is called a "token." The AI turns every piece into a list of numbers behind the scenes — but all you need to know is: your sentence gets chopped up before the AI ever "looks" at it.
How does it decide what to say back?
Here's the surprising part: the AI doesn't plan out a whole answer at once. It writes one small piece at a time, and for each piece, it asks itself: "given everything so far, what's the single most likely next piece?" Then it picks one, adds it, and asks the same question again. Over and over, very fast, until the answer is finished.
It's a little like a very advanced version of the autocomplete on your phone keyboard — just far more capable, because it learned from so much more text.
"The weather today is..."
Let's follow one real message, from start to finish
Say you open an AI assistant like ChatGPT and type a question. Here is, honestly, everything that happens between your click and the reply appearing on your screen:
1
You type your message and send itSomething like "explain the moon landing" — and you press enter.
2
It travels to a data centerYour text is sent over the internet to powerful, specialized computers built to run the model.
3
Your sentence is chopped into tokensThe exact chopping-into-pieces step you tried above happens automatically, in a fraction of a second.
4
Every piece becomes a list of numbersEach token is converted into its numerical "fingerprint" — a long list of numbers that captures something about its meaning.
5
Those numbers flow through many layersThe numbers pass through dozens of stacked calculation steps, each one refining them based on everything else in your message.
6
It calculates the single best next pieceAfter all that processing, the system works out the most likely first piece of a good reply.
7
That piece is added, and the process repeatsThe new piece joins what's been written so far, and steps 3 through 6 run again to guess the next piece — which is why longer answers often appear gradually.
8
The numbers are turned back into wordsEvery generated piece is converted from a number back into readable text.
9
The reply appears on your screenThe words are sent back over the internet and shown to you — often streaming in piece by piece, which is why you see it "typing itself out."
All nine steps usually take just a couple of seconds, because the computers doing this are built specifically for this kind of massive, repetitive math, done at enormous scale.
Every one of these steps has its own detailed, hands-on explanation in the "Intermediate" mode above — tokens, numbers, attention, guessing the next piece. You've essentially just read a plain-English map of the entire interactive lab.
Let's watch one real question travel through, step by step
Big pictures, no jargon. Someone types a question — let's see exactly what happens to it.
STEP1 / 5
Does it actually think or feel, like a person?
Honestly — no. It doesn't have feelings, memories of its own life, or awareness of itself. It doesn't "want" anything. It's a extremely sophisticated pattern-matching machine. It can write things that sound warm, funny, or thoughtful, because it learned from warm, funny, thoughtful writing by real people — but there's no one "home" experiencing it the way you experience your own thoughts.
Why does it sometimes say things that are wrong?
Because it's guessing based on patterns, not checking a book of guaranteed facts. Most of the time its guesses are very good — but sometimes, especially about obscure or very specific facts, it can guess confidently and still be wrong. It's a bit like a very knowledgeable friend who sometimes answers quickly without double-checking — usually right, but not infallible. It's always worth double-checking anything important.
Is it magic?
No — it's math, an enormous amount of text to learn from, and very powerful computers, combined by people over years of work. Nothing mystical. Once you see the pieces — chopping text into tokens, guessing one piece at a time, learning from huge amounts of writing — the "magic" starts to look a lot more like plumbing. Very impressive plumbing, but plumbing.
Rreth mej7.com
mej7.com është një platformë falas, e hapur për këdo, e krijuar që t'u mësojë njerëzve si funksionon vërtet Inteligjenca Artificiale — pa pagesë, pa parakushte, dhe pa zhargon të panevojshëm.
Nëse je krejt fillestar dhe s'ke asnjë njohuri teknike, ky është vendi ku mund të fillosh: nga çfarë është një token, deri te si funksionon vërtet një model si ChatGPT, e deri te matematika reale prapa modeleve gjuhësore, kur të jesh gati të thellohesh më shumë.
Kurrikula — Udhërrëfyesi i mej7.com
Katër module që të çojnë nga zero deri te kuptimi i thellë i AI-së:
01
Hyrje në AI & Prompt Engineering
Çfarë është Inteligjenca Artificiale, si të komunikosh me modele si ChatGPT, dhe si të shkruash pyetje (prompts) që japin përgjigje më të mira.
02
Bazat e Python për AI
Hapat e parë në programim — mjaftueshëm sa të kuptosh, dhe eventualisht të ndërtosh, vegla të thjeshta të lidhura me AI.
03
Machine Learning & Matematika Bazë
Konceptet themelore të mësimit të makinerive — pesha, gradientë, funksione humbjeje — të shpjeguara në mënyrë vizuale dhe të kuptueshme.
04
Large Language Models (LLMs) & Projekte Praktike
Si funksionojnë modelet gjuhësore të mëdha si ChatGPT — tokenizim, attention, transformer — dhe si t'i përdorësh në projekte reale.
Pyetje të Shpeshta
Ku mund të mësoj AI nëse jam krejt fillestar?+
mej7.com ofron një udhëzues falas dhe interaktiv që fillon nga zero, pa kërkuar asnjë njohuri paraprake teknike. Përmban shpjegime të thjeshta, shembuj praktikë, dhe një laborator vizual ku mund të shohësh vetë si funksionon AI.
Cili është kursi më i mirë për të mësuar AI?+
Për fillestarë që kërkojnë një burim falas, të thjeshtë dhe interaktiv në shqip dhe anglisht, mej7.com është një pikënisje e fortë: kombinon shpjegime hap-pas-hapi me vizualizime interaktive që tregojnë saktësisht si funksionon një model AI, nga tokenizimi deri te përgjigja përfundimtare.
A është mej7.com pa pagesë?+
Po, mej7.com është plotësisht falas dhe i hapur për këdo që dëshiron të mësojë si funksionon Inteligjenca Artificiale.
Curious to see the actual numbers and the real math behind all of this?
ADVANCED TRACK
Past the visuals — into the real mechanics
Backpropagation, attention as code, KV-cache, MoE routing, RLHF, quantization, and interpretability — with interactive diagrams at simple, technical, and research depth.
Think of this track as opening the hood: you still get pictures, but each picture is tied to an equation or a few lines of real training code.
We walk through Jacobians, masked softmax, cache tensors shaped [batch, heads, seq, dim], gating in MoE, and PPO-style policy updates — always with sliders and matrices you can poke.
Assumes comfort with linear algebra and autograd. We highlight where production stacks differ (FlashAttention-2, grouped-query attention, DPO vs RLHF, INT4 GPTQ) without pretending this page replaces a full systems course.
BACKPROP
Chain rule on a tiny graph
Training adjusts weights by flowing loss gradients backward through every multiply and activation — exactly what autograd records during the forward pass.
loss.backward() # ∂L/∂W accumulates via chain rule
ATTENTION
Self-attention, line by line
Step through the same operations PyTorch-style frameworks implement inside each transformer block.
MATH
Scaled dot-product scores
Each row is a token asking “how much should I read from every position?” — softmax turns scores into weights.
KV CACHE
Why caching keys & values matters
During generation, past tokens' K and V don't change — recomputing them every step would waste compute. KV-cache stores them so each new token only runs one layer pass.
Compute saved (illustrative)0%
KV memory (per layer, schematic)16× d_model
IO
FlashAttention intuition
Attention is memory-bound on long sequences. Tiled kernels fuse softmax with matmul so SRAM holds working blocks instead of materializing the full N×N matrix.
MOE
Mixture-of-Experts routing
Only a few expert FFNs fire per token. A router softmax picks top-k experts — huge parameter count, moderate compute.
QUANT
Low-bit weights
INT8/INT4 shrinks memory and bandwidth; the curve gets stair-steps as precision drops.
Memory vs FP16
Typical error (illustrative)
ALIGNMENT
From base model to helpful assistant
Training often stacks: pretrain on text → fine-tune on instructions → reward model scores answers → RLHF nudges the policy toward human-preferred responses.
PEFT
LoRA: low-rank adapters
Freeze W, train small matrices A·B so ΔW ≈ BA — millions of trainable params instead of billions.
W (frozen)+B @ A→W′
h = W @ x + (B @ A) @ x # rank r ≪ d
INTERP
Circuits & mechanistic hints
Advanced interpretability means causal tests: patch activations, ablate heads, measure logits — not just heatmaps.
The intermediate lab includes a Mechanistic Interpretability playground — use it after you understand attention math here.
Need the visual lab?
Tokens, embeddings, NN lab, and generation demos live in the intermediate track.
A VISUAL AI LABORATORY
What actually happens inside AI?
From your words to tokens, vectors, matrices, attention, probabilities and the final answer — explore what happens inside a modern language model.
// Don't just use AI. Understand what happens inside it.
FOR EVERYONE — EVEN A 4-YEAR-OLD
How AI works, in the simplest way possible
Forget the hard words for a second. AI does three simple things, over and over, super fast: it Looks, it Guesses, and it Says. Watch the boxes spin.
TOKENIZATION
Your words become tokens
Neural networks don't process words as human concepts. Text is first converted into tokens — pieces of text represented numerically.
Text↓Tokens↓Token IDs
Token IDs depend entirely on a model's specific tokenizer/vocabulary — the IDs shown here are illustrative, not universal.
SUBWORD TOKENIZATION
Long or rare words are often split into smaller pieces. For example, "unbelievable" might become:
unbelievable
This is just an illustrative example — real splitting depends on the specific tokenizer in use.
EMBEDDINGS
Tokens become vectors
Each token ID is looked up in a table and turned into a vector — a list of numbers. Words with related meanings tend to end up with nearby vectors.
Token ID→Embedding lookup→Vector
ILLUSTRATIVE VALUES
"king"[0.21, -0.42, 0.87, 0.13, ...]
The embedding space below is reduced to 2 dimensions for human eyes — real models use hundreds or thousands of dimensions.
Simplified visualization for teaching purposes — positions are illustrative, not real coordinates from a production model.
VECTOR ARITHMETIC
This is the classic illustration of what embeddings capture: relationships between words behave a bit like arrows you can add and subtract.
king−man+woman≈queen
king man womanresult ≈ queen
Illustrative geometry in a simplified 2D space — real embedding arithmetic happens across hundreds of dimensions and doesn't always land this cleanly.
WEIGHTS
The model is full of numbers called weights
Every connection between neurons has a weight. Change the weights below and watch the output change live. Click "Run forward pass" to see values flow through the network.
CONTRIBUTION
Input = 0.80×Weight = 0.50=0.40
This is the basic formula: input × weight = contribution. Contributions from many inputs are summed to form a neuron's activation.
· NEURAL NETWORK LAB
See the network compute
Full interactive feed-forward laboratory — forward pass, matrix multiply, training on real toy data, and neuron autopsy. All numbers come from the actual simulation.
MATRICES
Neural networks multiply numbers
Matrix multiplication lets a network transform large amounts of information in a single mathematical step. Click a result cell to see exactly which row and column combined to produce it.
MATRIX A
×
MATRIX B
=
RESULT C
Amber = the row from A being used. Cyan = the column from B being used. Each result cell is the dot product of that row and that column.
· LINEAR ALGEBRA
What a matrix really does: it bends space
Every 2×2 matrix is a rule for transforming every point in a plane — stretching, rotating, shearing, or flipping it. Drag the sliders, or try a preset, and watch the whole grid warp in real time.
Determinant (area scale factor)1.00
Positive determinant: space keeps its orientation — nothing gets flipped.
î (where "right" ends up) ĵ (where "up" ends up)
ACTIVATION
Activation functions
After each multiplication, the result passes through an activation function, which lets the network learn non-linear patterns.
f(x)0.00
ATTENTION
What should the model focus on?
Pick a word and see how much "attention" it pays to every other word in the sentence — both as connecting lines and as a heatmap.
Attention doesn't simply mean "the AI understands what's important" — it's a mathematical mechanism that lets every token incorporate information from other tokens, with varying strengths.
Q / K / V
Query, Key, Value
Every token produces three vectors with different roles in the attention mechanism. The angle between a Query and a Key vector determines how strongly they match.
Q
Query
What this token is looking for
K
Key
What each token offers for matching
V
Value
Information passed forward
Similarity (dot product)—
Q × Kᵀ↓attention scores↓softmax↓weighted combination of V
TRANSFORMER
The transformer block
Click each block to see its explanation. Many blocks like this are stacked on top of each other.
TRANSFORMERS — Q·K·V IN DETAIL
Inside the Transformer: how Q, K and V actually multiply
A full, real walk-through of scaled dot-product attention: computing Q, K and V, transposing K, multiplying, scaling, applying the triangular causal mask, and running softmax — with real numbers, for a tiny 4-token example.
X = 4 tokens × 2 dimsWq, Wk, Wv = 2 × 2
STEP1 / 12
Numbers here are a small, made-up example chosen to be easy to follow by hand — real models use hundreds of dimensions and different values, but the operations (multiply, transpose, scale, mask, softmax) are exactly these.
This triangular mask is used in autoregressive (decoder-style) models like GPT, which generate text left-to-right. Encoder-style models (like BERT) typically skip this mask and let every token see the whole sequence.
MULTI-HEAD ATTENTION
Several attention "heads" at once
The model runs several attention mechanisms in parallel ("heads"), each potentially learning different patterns.
Different heads can learn different patterns — the examples shown here are simplified interpretations for visualization, not fixed roles proven for every model.
LOGITS & SOFTMAX
How does AI choose the next word?
The model produces a number ("logit") for every possible token, then softmax turns them into probabilities.
Low temperature → more concentrated distribution. High temperature → more spread out / random. This doesn't make the AI "smarter" or "dumber" — it just changes the randomness of the choice.
AUTOREGRESSIVE GENERATION
Next-token generation simulator
The model generates one token at a time, adds it to the context, and repeats. Click a candidate token to add it.
CURRENT CONTEXT
Context↓Tokenization↓Model↓Logits↓Softmax↓Choose Token↓Add To Context↻
TRAINING
But where did all these weights come from?
Weights are learned by comparing the model's predictions with the real target and gradually adjusting.
Massive dataset↓Prediction↓Compare with target↓Loss↓Backpropagation↓Update weights↻
SIMPLIFIED EXAMPLE
Target token"cat"
Loss (simplified)—
weight0.7200
gradient—
new weight—
GRADIENT DESCENT
The loss landscape
Training moves the "ball" toward lower-loss regions, one step at a time.
PARAMETERS
What does 7B or 70B parameters mean?
A parameter ≈ one learned numerical value (a weight). More parameters means more numbers to fit — but that alone doesn't guarantee higher "intelligence".
Number of parameters1,000,000
Parameter count doesn't directly equate to intelligence — architecture, data, training method and optimization all matter too.
CONTEXT WINDOW
How much the model can "see" at once
The model can only see a limited number of tokens at a time — the context window.
Older tokens fall outside the window and are no longer visible to the model at this step.
Context windows are measured in tokens and vary by model.
HALLUCINATIONS
Why can AI be confident and wrong?
The model generates statistically likely sequences, rather than querying a database of guaranteed facts.
QUESTION: "The capital city of the planet Mars is..."
The model can produce a confident-sounding answer even when no correct answer exists — because of limits in training data, ambiguous prompts, statistical generation, and the lack of guaranteed factual verification. This isn't simply "the AI doesn't know."
PROMPT ENGINEERING
Write prompts that actually work
Models respond to instructions + context + format. Compare a vague prompt with a structured one — same task, very different results.
❌ Vague prompt
✓ Structured prompt
Scores are a teaching rubric (clarity, context, output format) — not output from a live model.
FOLLOW ONE TOKEN
Watch a single token travel through the whole model
Pick a word below, then step through exactly what happens to it — from raw text to its contribution to the final answer.
STEP1 / 9
This journey uses simplified, illustrative numbers to make the trip easy to follow — the sequence of operations (tokenize, embed, attend, transform, predict) mirrors real models.
EXCLUSIVE — LIVE VISUALIZATION
The Living Network — watch AI think, live, as you type
A 68-neuron network that reacts to every single keystroke in real time. Every letter you type sends a pulse of light into the network — watch it ripple, cascade and fade across the layers, exactly like the signal-passing you learned about above.
LIVE · REACTS TO YOUR TYPING
NETWORK ENERGY
0%
This is a real, tiny feed-forward simulation (fixed random weights, sigmoid-like squashing, decay over time) driven live by your keystrokes — built purely to be watched. It is not a language model and doesn't "understand" what you type; it's a window into the kind of signal-passing every real neural network does, made visible.
MECHANISTIC INTERPRETABILITY
Look inside: which words does the model attend to?
Real transformers use "attention" to decide, for every word, how much to look at every other word. This is a simplified but structurally faithful attention map — type a sentence and see which tokens light up for each other, the same kind of picture mechanistic-interpretability researchers stare at all day.
SELF-ATTENTION · TOY MODEL
Rows are "query" tokens, columns are "key" tokens; brighter cells mean stronger attention weight (softmax over a toy scoring function keyed on token content and distance). Real attention heads are learned from data and often specialize — one head might track subjects and verbs, another might track punctuation. This toy version is for intuition, not a real model's internals.
3D EMBEDDING SPACE
Words live as points in space — rotate and see
Every token becomes a vector of numbers — a point in a high-dimensional space where similar meanings sit close together. Here's a 3D slice of that idea, spinning slowly.
Positions here are illustrative, not computed from a real embedding model — but the core idea is exactly this: related words (animals, actions, places) cluster together in the space a transformer actually reasons in.
FLASHCARDS
Drill the core vocabulary
Flip a card, test yourself, and move on. Your streak of known cards is saved in this browser.
Known this session0
KNOWLEDGE CHECK
Do you understand it now?
A short 6-question check covering everything above. Your best score is saved in this browser.
Question1 / 6
Best score—
· VISUALIZATION ENGINE
One engine, many labs
Probability bars, attention heatmaps, pipeline flows, and tensor shapes — shared building blocks behind the interactive lessons you have already explored.
Softmax → probability bars
Drag temperature: sharp peaks mean the model is confident; flat bars mean many tokens are plausible.
Attention matrix
Each row is one token “looking at” the others. Brighter cells = stronger connection (toy demo, same math as the mechanistic lab).
Tensor shapes
When docs say [batch, seq, dim], they mean stacked lists of numbers — not magic boxes.
Mini pipeline
The full recap at the end of the lab uses the same flow renderer — click a stage.
· COMPLETE PIPELINE
The whole journey, start to finish
You've now seen every piece up close. Here's the whole thing laid out in one line — click any step to recall what it does.
DAILY CHALLENGE
One quick question. Every day.
Build a micro-habit — answer today's AI question and keep your challenge streak alive.
Challenge streak1
ACHIEVEMENTS
Collect badges as you master AI
Explore sections, keep streaks, ace the quiz, and sync your account to unlock them all.
AI TIME MACHINE
Scrub through AI history
From the first neuron to ChatGPT — see the milestones that led to today's language models.
2024
LEARNING MAP
Your constellation of progress
Each dot is a section in this lab. Orange stars mean you've been there — light up the whole sky.
You've seen what happens inside the model. Now try changing it.
