Tendril

Tendril · Explorers · AI Foundations

How AI Read Almost the Whole Internet

AI learned by reading a huge pile of books, websites, and writing.

40 min · Reviewed 2026

The big idea

Before an AI chatbot can help you, it has to learn. To learn, it reads — way more than any human ever could. We're talking billions of webpages, books, and articles. It's the world's biggest library, and the AI has 'read' most of it.

Some examples

AI has read Wikipedia, lots of books, and tons of websites.
It also reads forum posts, recipes, news, and how-to guides.
All that reading is called 'training data'.
The AI doesn't remember the exact pages — it remembers patterns from them.

Try it!

Ask an AI 'What's the recipe for chocolate chip cookies?' It knows the answer because it read thousands of cookie recipes. Now ask it your name. It probably doesn't know — because your name isn't in the giant library.

AI Is Only as Smart as What It Read

The big idea

There's a saying with computers: 'garbage in, garbage out.' If AI was trained on a pile of wrong info, it'll repeat that wrong info. The quality of what AI READ becomes the quality of what AI SAYS.

Some examples

If AI learned from old books, it'll give old-fashioned answers.
If AI saw lots of jokes, it might be funnier.
If AI read mostly English, it'll be best in English.
Bad data = bad answers, even from a fancy AI.

Try it!

Ask AI a question about something brand new from this week. Does it actually know it?

AI Is a Super-Smart Copy Machine

The big idea

Did you know AI is kind of like a giant copy machine that learned to remix? It read tons of human writing and learned patterns. It doesn't invent — it shuffles what people already made.

Some examples

AI writes poems by mixing styles from real poems.
AI draws cats using millions of cat photos it studied.
Brand-new ideas are very rare for AI.
Humans came up with everything AI knows first.

Try it!

Ask AI for a 'totally brand new' joke. Then check — is it really new, or is it borrowed?

The Snack Bowl of Examples That Trains an AI

The big idea

Imagine a giant bowl filled with millions of pictures, sentences, and sounds. AI scoops out examples and slowly learns from them. The bigger and better the bowl, the smarter the AI gets.

Some examples

A photo bowl full of cats so AI learns what cats look like.
A book bowl full of stories so AI learns how sentences flow.
A song bowl so AI can suggest music you might like.
A drawing bowl so AI can guess what your doodle is.

Try it!

If you wanted an AI to draw funny cartoon cats, what 10 things would you put in its training bowl? Make a list with a grown-up.

Where AI Learned Everything It Knows

The big idea

AI learned to write by reading billions of pages — websites, books, and online posts — long before it ever talked to you.

Some examples

AI has 'read' Wikipedia, news sites, and lots of books.
If something happened after AI's training, AI may not know about it.
AI picked up grammar and facts the same way: by seeing them many times.
Different AI models read different piles of text.

Try it!

Ask an AI 'What's the latest news?' — most can't tell you, because their reading time ended months or years ago. Now you know why!

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-explorers-foundations-AI-and-the-big-library

What is the core idea behind "How AI Read Almost the Whole Internet"?
1. AI learned by reading a huge pile of books, websites, and writing.
2. knowing vs feeling
3. Eliminate the need to update the eval set as the product evolves
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which term best describes a foundational idea in "How AI Read Almost the Whole Internet"?
1. internet text
2. training data
3. data scale
4. knowing vs feeling
A learner studying How AI Read Almost the Whole Internet would need to understand which concept?
1. training data
2. data scale
3. internet text
4. knowing vs feeling
Which of these is directly relevant to How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. knowing vs feeling
4. data scale
Which of the following is a key point about How AI Read Almost the Whole Internet?
1. AI has read Wikipedia, lots of books, and tons of websites.
2. It also reads forum posts, recipes, news, and how-to guides.
3. All that reading is called 'training data'.
4. The AI doesn't remember the exact pages — it remembers patterns from them.
Which of these does NOT belong in a discussion of How AI Read Almost the Whole Internet?
1. All that reading is called 'training data'.
2. It also reads forum posts, recipes, news, and how-to guides.
3. AI has read Wikipedia, lots of books, and tons of websites.
4. knowing vs feeling
What is the key insight about "The rule" in the context of How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. AI learns from a giant pile of human writing.
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which statement accurately describes an aspect of How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Compare standard attention to FlashAttention v2 and v3 at conceptual level
4. Before an AI chatbot can help you, it has to learn. To learn, it reads — way more than any human ever could.
What does working with How AI Read Almost the Whole Internet typically involve?
1. Ask an AI 'What's the recipe for chocolate chip cookies?' It knows the answer because it read thousands of cookie recipes.
2. knowing vs feeling
3. Eliminate the need to update the eval set as the product evolves
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which best describes the scope of "How AI Read Almost the Whole Internet"?
1. It is unrelated to foundations workflows
2. It focuses on AI learned by reading a huge pile of books, websites, and writing.
3. It applies only to the opposite professional tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Some examples
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which section heading best belongs in a lesson about How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Compare standard attention to FlashAttention v2 and v3 at conceptual level
4. Try it!
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling

← Back to interactive lesson

Tendril · Explorers · AI Foundations

How AI Read Almost the Whole Internet

AI learned by reading a huge pile of books, websites, and writing.

40 min · Reviewed 2026

The big idea

Some examples

AI has read Wikipedia, lots of books, and tons of websites.
It also reads forum posts, recipes, news, and how-to guides.
All that reading is called 'training data'.
The AI doesn't remember the exact pages — it remembers patterns from them.

Try it!

AI Is Only as Smart as What It Read

The big idea

There's a saying with computers: 'garbage in, garbage out.' If AI was trained on a pile of wrong info, it'll repeat that wrong info. The quality of what AI READ becomes the quality of what AI SAYS.

Some examples

If AI learned from old books, it'll give old-fashioned answers.
If AI saw lots of jokes, it might be funnier.
If AI read mostly English, it'll be best in English.
Bad data = bad answers, even from a fancy AI.

Try it!

Ask AI a question about something brand new from this week. Does it actually know it?

AI Is a Super-Smart Copy Machine

The big idea

Did you know AI is kind of like a giant copy machine that learned to remix? It read tons of human writing and learned patterns. It doesn't invent — it shuffles what people already made.

Some examples

AI writes poems by mixing styles from real poems.
AI draws cats using millions of cat photos it studied.
Brand-new ideas are very rare for AI.
Humans came up with everything AI knows first.

Try it!

Ask AI for a 'totally brand new' joke. Then check — is it really new, or is it borrowed?

The Snack Bowl of Examples That Trains an AI

The big idea

Imagine a giant bowl filled with millions of pictures, sentences, and sounds. AI scoops out examples and slowly learns from them. The bigger and better the bowl, the smarter the AI gets.

Some examples

A photo bowl full of cats so AI learns what cats look like.
A book bowl full of stories so AI learns how sentences flow.
A song bowl so AI can suggest music you might like.
A drawing bowl so AI can guess what your doodle is.

Try it!

If you wanted an AI to draw funny cartoon cats, what 10 things would you put in its training bowl? Make a list with a grown-up.

Where AI Learned Everything It Knows

The big idea

AI learned to write by reading billions of pages — websites, books, and online posts — long before it ever talked to you.

Some examples

AI has 'read' Wikipedia, news sites, and lots of books.
If something happened after AI's training, AI may not know about it.
AI picked up grammar and facts the same way: by seeing them many times.
Different AI models read different piles of text.

Try it!

Ask an AI 'What's the latest news?' — most can't tell you, because their reading time ended months or years ago. Now you know why!

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-explorers-foundations-AI-and-the-big-library

What is the core idea behind "How AI Read Almost the Whole Internet"?
1. AI learned by reading a huge pile of books, websites, and writing.
2. knowing vs feeling
3. Eliminate the need to update the eval set as the product evolves
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which term best describes a foundational idea in "How AI Read Almost the Whole Internet"?
1. internet text
2. training data
3. data scale
4. knowing vs feeling
A learner studying How AI Read Almost the Whole Internet would need to understand which concept?
1. training data
2. data scale
3. internet text
4. knowing vs feeling
Which of these is directly relevant to How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. knowing vs feeling
4. data scale
Which of the following is a key point about How AI Read Almost the Whole Internet?
1. AI has read Wikipedia, lots of books, and tons of websites.
2. It also reads forum posts, recipes, news, and how-to guides.
3. All that reading is called 'training data'.
4. The AI doesn't remember the exact pages — it remembers patterns from them.
Which of these does NOT belong in a discussion of How AI Read Almost the Whole Internet?
1. All that reading is called 'training data'.
2. It also reads forum posts, recipes, news, and how-to guides.
3. AI has read Wikipedia, lots of books, and tons of websites.
4. knowing vs feeling
What is the key insight about "The rule" in the context of How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. AI learns from a giant pile of human writing.
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which statement accurately describes an aspect of How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Compare standard attention to FlashAttention v2 and v3 at conceptual level
4. Before an AI chatbot can help you, it has to learn. To learn, it reads — way more than any human ever could.
What does working with How AI Read Almost the Whole Internet typically involve?
1. Ask an AI 'What's the recipe for chocolate chip cookies?' It knows the answer because it read thousands of cookie recipes.
2. knowing vs feeling
3. Eliminate the need to update the eval set as the product evolves
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which best describes the scope of "How AI Read Almost the Whole Internet"?
1. It is unrelated to foundations workflows
2. It focuses on AI learned by reading a huge pile of books, websites, and writing.
3. It applies only to the opposite professional tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Some examples
4. Compare standard attention to FlashAttention v2 and v3 at conceptual level
Which section heading best belongs in a lesson about How AI Read Almost the Whole Internet?
1. knowing vs feeling
2. Eliminate the need to update the eval set as the product evolves
3. Compare standard attention to FlashAttention v2 and v3 at conceptual level
4. Try it!
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling
Which of the following is a concept covered in How AI Read Almost the Whole Internet?
1. training data
2. internet text
3. data scale
4. knowing vs feeling

← Back to interactive lesson