Skip to content

The full curriculum

11 lessons live now. 59 exercises. Every concept taught through a live playground, not slides.

26 free exercises
42 paid exercises
₹499 full access

Concepts

39 exercises
01

Hello, Intelligence

Free

Your first conversation with an LLM. Autocomplete, knowledge retrieval, creativity, and code generation.

Autocomplete Knowledge retrieval Hallucinations Creative generation Code completion
16 ex
02

The Currency of Tokens

Free

How LLMs read and write in chunks. Token limits, costs, and the difference between hard and soft constraints.

Tokenization Max output tokens Soft vs hard limits Format cost System instructions intro
10 ex
17

Hallucinations & Sycophancy

When the model lies confidently, agrees with your delusions, and validates your worst ideas.

Tokenization traps Sycophancy Confabulation The yes-man problem Emotional validation
7 ex
18

Safety, Ethics & Alignment

Guardrails, jailbreaking, implicit bias, and the values embedded in AI models.

RLHF Safety filters Jailbreaking Bias detection Value alignment
6 ex

Engineering

20 exercises
03

Temperature: Consistency

Using low temperature for deterministic, reliable outputs. Data extraction, classification, math.

Temperature 0.0 Deterministic output Data extraction
2 ex
04

Temperature: Creativity

Using high temperature for brainstorming, storytelling, and creative diversity.

High temperature Creative generation Output diversity
2 ex
07

System Instructions

The constitution of your model. Personas, formatting, guardrails, and scope control.

Personas Output formatting Scope guardrails Competitor blocking JSON/HTML output
7 ex
08

Zero-Shot Prompting

Asking the model to perform tasks without examples. When it works and when it doesn't.

Zero-shot classification Direct task prompting
2 ex
09

Few-Shot Prompting

Teaching the model by example. The single most powerful prompting technique.

Few-shot learning Pattern matching Style transfer Format enforcement
3 ex
10

Chain of Thought

Making models reason step by step. The trick that fixes most logic failures.

Step-by-step reasoning Scratchpad technique Logic improvement
2 ex
11

Stop Sequences

Controlling when generation ends. Preventing the model from speaking for the user.

Stop sequences Generation termination Dialog control
2 ex

Advanced

8 exercises
05

Top-K Sampling

Coming Soon

Limiting the selection pool of tokens. Greedy decoding vs wide-net generation.

Top-K Greedy decoding Token selection
2 ex
06

Top-P (Nucleus Sampling)

Coming Soon

Dynamic vocabulary control using cumulative probability.

Top-P Nucleus sampling Probability mass
1 ex
12

Formatting via Prompting

Coming Soon

Getting CSV, XML, Markdown tables, and structured output without special configs.

CSV generation Markdown tables Structured output
2 ex
13

Native JSON Mode

Coming Soon

Guaranteed valid JSON output using responseMimeType. Essential for developers.

JSON mode responseMimeType Structured API output
1 ex
15

Vision: Description

Coming Soon

Multimodal input. Upload an image and have the model describe what it sees.

Image input Multimodal models Visual description
1 ex
16

Vision: Reasoning

Coming Soon

Going beyond description. Asking the model to reason about image context.

Visual reasoning Context inference Image QA
1 ex

Coding

1 exercises
14

Coding Specialist

Coming Soon

Using specialized models for code generation. React, Python, debugging.

Code generation Specialized models Thinking budget
1 ex

On the roadmap

These lessons are in development. Buy now and get them automatically when they ship — no extra charge.

05 Top-K Sampling 2 ex
06 Top-P (Nucleus Sampling) 1 ex
12 Formatting via Prompting 2 ex
13 Native JSON Mode 1 ex
14 Coding Specialist 1 ex
15 Vision: Description 1 ex
16 Vision: Reasoning 1 ex

See something you want to learn?

The first 26 exercises are free. Open the playground and try Lesson 1.

Try Lesson 1 — Free

Stop watching. Start tinkering.