Cheatsheets

Full-stack + AI skills

A modern reference across frontend, backend, data, DevOps, and AI/LLM engineering.

47 entries

Frontend7

SSR / SSG / ISR

Server-render, static-generate, or incrementally revalidate

React Server Components

Render on the server, ship less client JS

Hydration

Attach interactivity to server-rendered HTML

Core Web Vitals

LCP, CLS, INP — Google's UX performance metrics

Accessibility (a11y)

Semantic HTML, ARIA, keyboard nav, contrast

Code splitting

Lazy-load chunks so the initial bundle stays small

Optimistic UI

Update the UI before the server confirms

Backend & APIs7

REST vs GraphQL

Resource endpoints vs a single typed query graph

JWT / OAuth 2.0

Stateless tokens / delegated authorization

Idempotency

Safe to retry — same request, same effect

Rate limiting

Throttle clients (token bucket, sliding window)

Webhooks

Server-to-server event callbacks (verify signatures!)

Message queues

Decouple work async (SQS, Kafka, RabbitMQ)

Caching layers

CDN, Redis, HTTP cache headers, memoization

Data & storage6

Indexes

Speed up reads; cost writes & storage

ACID transactions

Atomic, consistent, isolated, durable

Normalization

Reduce duplication vs denormalize for read speed

Migrations

Versioned, repeatable schema changes

Vector database

Store embeddings for semantic / similarity search

N+1 queries

Common ORM perf bug — batch or join instead

DevOps & cloud6

CI/CD

Automated build, test, deploy on every change

Containers

Package app + deps; run anywhere (Docker)

Infrastructure as Code

Declarative infra (Terraform, Pulumi)

Observability

Logs, metrics, traces — know what prod is doing

Blue-green / canary

Release strategies that limit blast radius

Serverless / edge

Run functions on-demand, close to users

AI / LLM basics6

Tokens

Sub-word units; billing & limits are per-token

Context window

Max tokens the model can attend to at once

Temperature

Sampling randomness — lower = more deterministic

Embeddings

Vectors capturing meaning for search & clustering

Streaming

Emit tokens as generated for responsive UX

Prompt vs fine-tune

Steer via prompt first; fine-tune for scale/format

AI engineering9

System prompt

Sets the model's role, rules, and tone

Few-shot prompting

Include examples to steer the output

Chain-of-thought

Ask the model to reason step by step

RAG

Retrieval-Augmented Generation — ground answers in fetched docs

Tool / function calling

Let the model invoke typed functions/APIs

Agents

LLM loops that plan, call tools, and act

MCP

Model Context Protocol — connect tools & data to LLMs

Evals & guardrails

Test prompt quality; constrain unsafe output

const r = await fetch("https://api.anthropic.com/v1/messages", {
  method: "POST",
  headers: {
    "x-api-key": KEY,
    "anthropic-version": "2023-06-01",
    "content-type": "application/json",
  },
  body: JSON.stringify({
    model: "claude-sonnet-4-6",
    max_tokens: 1024,
    messages: [{ role: "user", content: "Hi" }],
  }),
});

Call the Anthropic Messages API

Quality & practices6

Unit / integration / e2e

Test pyramid — many fast, few slow

Type safety

Catch errors at compile time (TypeScript)

Feature flags

Ship dark, roll out gradually, kill switch

12-factor app

Config in env, stateless processes, logs as streams

Least privilege

Grant the minimum access needed

Graceful degradation

Keep core features working when parts fail