AI Cheat Sheets

One-page AI cheat sheets for developers: how each mechanism works, the rules that hold, and when to use it. Click a sheet to watch it build step by step.

  • Home /
  • AI Cheat Sheets

Retrieval-Augmented Agents cheat sheet

The agent decides when to search again; without a retry cap and a grading step it loops or answers from garbage.

#AI Agent ArchitectureAnimated

→ Topic: Retrieval-Augmented Agents

Retrieval-Augmented Generation cheat sheet

Good retrieval is necessary, not sufficient: generation can still ignore the right document.

#Retrieval-Augmented GenerationAnimated

→ Topic: Retrieval-Augmented Generation

Reward Model Architecture cheat sheet

The score only ranks answers, and the policy compounds every bias in it until it learns to game the score.

#LLM Training & Pre-TrainingAnimated

→ Topic: Reward Model Architecture

RLHF cheat sheet

RLHF matches what annotators preferred, so a rising reward score can still hide a worse model.

#LLM Training & Pre-TrainingAnimated

→ Topic: RLHF

Role Prompting cheat sheet

A role changes how the model sounds and what it focuses on, never what it knows.

#Prompt EngineeringAnimated

→ Topic: Role Prompting

Scaling Laws cheat sheet

Scaling laws forecast loss before a run starts; balancing data with size beats adding parameters alone.

#LLM Training & Pre-TrainingAnimated

→ Topic: Scaling Laws

Sentence Transformers cheat sheet

Contrastive training, not the base model, is what makes sentence vectors comparable by meaning.

#Retrieval-Augmented GenerationAnimated

→ Topic: Sentence Transformers

Similarity Search Algorithms cheat sheet

Approximate search gives up a little accuracy for speed; weak results usually start in the embeddings, not the index.

#Retrieval-Augmented GenerationAnimated

→ Topic: Similarity Search Algorithms

Sparse Retrieval cheat sheet

BM25 and learned sparse models share one index and its limits; they differ only in who decides the weights.

#Retrieval-Augmented GenerationAnimated

→ Topic: Sparse Retrieval

State Space Model cheat sheet

Cost per token stays flat at any length, because the past is compressed into one state instead of compared token by token.

#Transformer & Attention InternalsAnimated

→ Topic: State Space Model

Structured Output Prompting cheat sheet

A schema can force well-formed JSON, yet a confident wrong value still passes, so validate values as well as shape.

#Prompt Ops & SecurityAnimated

→ Topic: Structured Output Prompting

SWE-bench cheat sheet

A pass on hidden tests scores a curated set of open-source issues, not your own codebase.

#Model EvaluationAnimated

→ Topic: SWE-bench

Synthetic Data Ethics cheat sheet

A synthetic record only protects people once a privacy attack on the set fails to trace it back.

#Synthetic Data & GenerationAnimated

→ Topic: Synthetic Data Ethics

Synthetic Data Generation cheat sheet

Generated rows stand in for real ones only after a model trained on them holds up on real inputs.

#Synthetic Data & GenerationAnimated

→ Topic: Synthetic Data Generation

System Prompts cheat sheet

A system prompt steers every reply, but critical rules still need checks outside the model.

#Prompt EngineeringAnimated

→ Topic: System Prompts

Temperature and Sampling cheat sheet

Temperature only changes which words get picked from the same probabilities; the model's knowledge stays the same.

#Inference OptimizationAnimated

→ Topic: Temperature and Sampling

Text-to-3D cheat sheet

Generated meshes are drafts: plan a retopology and UV repair pass before an engine or a rig accepts them.

#AI Audio, Video & 3DAnimated

→ Topic: Text-to-3D

Text-to-Speech cheat sheet

Neural speech is generated from scratch in stages; pick the engine by latency and deployment needs before comparing how voices sound.

#AI Audio, Video & 3DAnimated

→ Topic: Text-to-Speech

Tokenizer Architecture cheat sheet

A model only ever sees token IDs; the subword algorithm decides how text splits, and changing it later means retraining.

#Transformer & Attention InternalsAnimated

→ Topic: Tokenizer Architecture

Tool Use in Prompts cheat sheet

The model only proposes a call; your application executes it, and a valid schema does not make the choice right.

#Prompt Ops & SecurityAnimated

→ Topic: Tool Use in Prompts

Toxicity and Safety Evaluation cheat sheet

Benchmarks only see known attacks, so a safety check repeats after every model or prompt change.

#Safety & Red TeamingAnimated

→ Topic: Toxicity and Safety Evaluation

Training Data Quality cheat sheet

The data sets the ceiling; a hand-checked sample shows how far below it your labels sit.

#Training Data Quality & CurationAnimated

→ Topic: Training Data Quality

Transformer Architecture cheat sheet

Positional encoding restores word order, attention links every token at once, and the same block repeats layer after layer.

#Transformer & Attention InternalsAnimated

→ Topic: Transformer Architecture

Tree of Thoughts cheat sheet

Branching beats a single chain only when the evaluator can tell a good partial step from a bad one.

#Prompt EngineeringAnimated

→ Topic: Tree of Thoughts

Variational Autoencoder cheat sheet

Sampling from a learned distribution, made trainable by moving the randomness into a noise term, is what lets it generate.

#Neural Network ArchitecturesAnimated

→ Topic: Variational Autoencoder

Vector Indexing cheat sheet

An index trades a little recall for speed; tune it against an exact baseline so you know what it skips.

#Retrieval-Augmented GenerationAnimated

→ Topic: Vector Indexing

Vibe Coding cheat sheet

The agent writes the code, but context, tests and review decide whether it can ship.

#Agentic & Autonomous CodingAnimated

→ Topic: Vibe Coding

Vision Transformer cheat sheet

Self-attention over patches drops the built-in assumptions of convolution, so data or a pretrained backbone has to supply them.

#Transformer & Attention InternalsAnimated

→ Topic: Vision Transformer

Voice Cloning cheat sheet

A few seconds of audio give a speaker embedding; the base model still sets the quality ceiling, and nothing in it checks consent.

#AI Audio, Video & 3DAnimated

→ Topic: Voice Cloning

What is a computer use agent? cheat sheet

A computer use agent is an AI model that uses a browser or desktop like a person. See how it sees a page, why it is slow, and what is safe to hand over.

#AI Agent Architecture

→ Read the article