Hands-on track · Instructions that age with the stack

Great prompts
aren't written.
They're engineered.

PromptSq² is our interactive foundation—short lessons that make you fluent in the instruction layer every agentic stack depends on, so when the ecosystem shifts, you are steering the model with intent, not guesswork.

Skim. Learn. Ship.

Prompt engineering finally clicks — the moment you can see it.

Watch every word get tokenized, vectorized, and analyzed — see exactly what the model sees.

prompt.inputready

▸

0 chars · 0 tokenspromptforge v1.0

model.outputgenerating…

▸

In the silent depths of the library, a hidden door opened to the 1820s, revealing a corridor that felt deeply unnatural. Portraits lined the walls, their eyes following him — expressions shifting, as if they recognized him. At the far end, an open book displayed a list of names, all violently scratched out… except his. As he drew closer, the ink began to writhe, carving his name again and again across the page. Behind him, the door sealed shut—and from the portraits, something started to emerge.

output will appear here…

Under the hood

See what an LLM actually does with your prompt

The atoms of AI language

Tokens

Every word in a prompt gets shattered into fragments called tokens — the only language a model truly speaks. See how a sentence dissolves and reforms as raw machine currency.

Tokens

Tokens 0Characters 0

Tokens appear as you type…

Your words, mapped in space

Embeddings

LLMs don't read words — they navigate a cosmos of numbers. Every token becomes an embedding: a precise point in a high-dimensional mathematical universe where meaning lives as geometry. See how words become vectors.

embeddings.2d

Type to plot vectors…

Eight minds reading a prompt at once

Attention Heads

When an AI reads your prompt, it doesn't go word by word—it looks for connections. Each word asks: Which other words matter most to me ? Every attention head answers that differently. One links "door" to "library." Another connects "1820s" to the unfolding scene. A third ties "mysterious" to "event." Together, they form a web of relationships—revealing what your prompt truly means, not just what it says.

attention-head-view

LayerAttention

hover a token on the left to inspect its attentionWrite→ attending

Not all words pull their weight

Word Weightage

Not every word pulls equal influence. This view separates the high-signal keywords from semantic filler — so you can see exactly which words are doing the heavy lifting and which ones are just noise.

word-weight.analysisdominant: —

Type to measure semantic weight…

Every word costs something

Token Economy

Every token has a cost in compute and in the model's limited context window. This view shows how token usage adds up — and why writing with precision, not padding, is the hallmark of a skilled prompt engineer.

token-economy.log

0tokens

$0 total($0.0001/token)

Short lessons · Big payoff

Most courses talk at you.
We show you.

In a fast-moving agentic AI world, you need lessons you can skim between meetings—not a wall of text pretending to be a curriculum. We built short, visual modules where the concept does the explaining, not a paragraph you'll forget by tomorrow.

Every other course

PromptSq2

✕Dense paragraphs you skim and forget

✓Visual, animated lessons that make concepts click

✕Memorize jargon for a quiz

✓Understand the idea by seeing it shown

✕Lecture slides that go in one ear, out the other

✓Short, sharp lessons built around one idea at a time

✕Forgotten in a week

✓Crisp lessons, not long reads — so you retain more.

No walls of text

Every lesson is short, focused, and visual. We stripped out the filler so you spend zero time re-reading sentences — and all of it actually learning.

Show, don't tell

Tokenization, embeddings, attention — we don't define these. We show them. Animated, visual, in plain language. If you can see it, you'll understand it.

Fun by design

Learning feels slow when it's boring. We made it genuinely enjoyable — the kind of course you actually want to open, not the kind you drag yourself through.

Finally, a fun way to learn prompt engineering.

Join engineers, founders, and builders who finally got it — not from a textbook, but from a course that actually made sense.

Start learning for free →

See the full curriculum

No credit card. No fluff. Just the real thing.

How everyone else learns

Trial & Error

Copy-paste prompts from Reddit threads
Tweak random words and hope for better output
No idea why one prompt works and another doesn't
Stuck repeating the same beginner mistakes

How PromptSq² students learn

Predictive Engineering

See exactly how the model tokenizes your words
Understand which parts of your prompt the model ignores
Identify high-signal keywords before you even run the model
Build mental models that work across every AI platform

60-second reality check

Vague instructions don't confuse the model magically — they make it confidently wrong. Watch a chat spiral when context never shows up.

See the chat go wrong →

Full Course — Coming Soon

From curious to certified.

Structured modules, live labs, and a certification that proves you're not just using AI — you understand it.

Structured modules from beginner to advanced
Hands-on labs with live AI visualisation
Multi-model A/B prompt testing
Context compression & chain-of-thought mastery
Course completion certificate

Built for builders who can't wait months to learn what shipped last week.

Great promptsaren't written.They're engineered.

See what an LLM actually does with your prompt

Tokens

Embeddings

Attention Heads

Word Weightage

Token Economy

Most courses talk at you.We show you.

No walls of text

Show, don't tell

Fun by design

Trial & Error

Predictive Engineering

From curious to certified.

See what an LLM actually does with your prompt

Tokens

Embeddings

Attention Heads

Word Weightage

Token Economy

Most courses talk at you.We show you.

No walls of text

Show, don't tell

Fun by design

Great prompts
aren't written.
They're engineered.

Most courses talk at you.
We show you.

Most courses talk at you.
We show you.