Guided paths
Learn
Four guided paths from beginner to advanced, built around tutorials, guides, and tools.
- 1Set up your stack
The modern AI app stack: Next.js, Vercel AI SDK, and the tools that tie them together.
- 2Learn the AI SDK
Streaming responses, tool calling, and structured output in TypeScript.
- 3Build an AI agent
Go beyond chat. Build agents that reason, plan, and take action.
- 4Understand RAG
Retrieval-augmented generation: give your AI app access to real data.
- 5Take the course
Structured lessons with code examples, quizzes, and projects.
- 6Explore tools
Browse the full AI tools directory to find what fits your stack.
- 1Pick your tools
A ranked breakdown of the 10 best AI coding tools available right now.
- 2Learn Claude Code
The complete guide to Claude Code: setup, usage, and real workflows.
- 3Power user tips
Advanced techniques, shortcuts, and patterns for daily use.
- 4Set up MCP
Connect Claude Code to external data sources with MCP servers.
- 5Vibe coding
The art of building software by describing what you want in natural language.
- 6Compare options
Side-by-side comparisons of AI coding tools, frameworks, and models.
- 1Understand agents
What AI agents are, how they work, and why they matter for developers.
- 2Build with TypeScript
Hands-on guide to building your first agent with tool use and reasoning.
- 3Add MCP servers
Give your agents access to databases, APIs, and file systems via MCP.
- 4Multi-agent systems
Orchestrate multiple agents that collaborate, delegate, and coordinate.
- 5Explore frameworks
Browse AI frameworks, SDKs, and agent toolkits in the tools directory.
- 1Learn agentic coding
Build the operating model for terminal agents, worktrees, delegation, and code review.
- 2Study Codex automation patterns
Use Codex for sandboxed execution, headless checks, CI workflows, and batch repo operations.
- 3Compare coding agents
Pick where Codex, Claude Code, Cursor, and Copilot fit in your workflow.
- 4Add review discipline
Use the Claude Code course modules on CI, security, costs, and governance as the review baseline.
- 5Choose supporting tools
Find issue trackers, observability tools, local models, and agent frameworks for recurring workflows.
- 1What is Claude Code?
Start here. An intro to the most popular AI coding agent.
- 2What is MCP?
Model Context Protocol explained simply, with practical examples.
- 3Pricing guide
What each tool costs, free tiers, and how to pick the right plan.
- 4Free tools
48 free developer tools you can use right now, no signup required.
- 5Glossary
Look up any AI development term. Plain-language definitions.
Shipping Features with AI Agents
Developers who want an agent to carry a feature from idea to merged PR
- 1Get oriented on the stack
The tools that make up an agentic dev workflow in 2026, and how they fit together.
- 2Build your first agent
Hands-on guide to building an agent that can plan and take action, not just chat.
- 3Move fast without losing the plot
Ship features by describing intent, while keeping enough structure to stay in control.
- 4Harden it for production
The production principles that keep an agent-shipped feature reliable after launch.
- 5Set review rules before you ship
The new rules for reviewing agent-authored pull requests before they merge.
- 1Debug agent workflows
Where agent-driven bugs actually come from, and how to trace them.
- 2Handle errors reliably
Error handling patterns that stop a flaky API call from becoming a flaky feature.
- 3Automate test generation
Which AI test generation tools actually catch bugs, compared head to head.
- 4Add AI code review
CodeRabbit, DeepSource, and Greptile compared, so review catches what tests miss.
- 5Manage the review bottleneck
Why AI coding agents move the bottleneck to review queues, and how to keep up.
- 1See what everything costs
A full pricing comparison across AI coding tools, free tiers included.
- 2Compare the two leaders
Claude vs GPT for coding: which model writes better TypeScript.
- 3Understand the new tiers
What the newest GPT and Claude tiers mean for picking a coding model.
- 4Weigh cost against quality
The cost-quality gap between Fable 5 and DeepSeek V4, measured on real tasks.
- 5Plan around usage-based billing
What changed with Copilot's usage-based billing, and what it costs now.
Hands-On Tutorials
Pick a skill, then a tool. Each tutorial is a working walkthrough with code, common pitfalls, and what to build next.
Build a RAG pipeline
RAGLangChain
LangChain gives you battle-tested abstractions for retrieval-augmented generation: document loaders, text splitters, vector stores, and retrievers. This walkthrough builds a working RAG pipeline in TypeScript that ingests your docs and answers questions over them with citations.
LlamaIndex
LlamaIndex is the data framework for LLM apps. It excels at structured ingestion: parsing PDFs, websites, Notion, and SQL into a queryable index. This guide builds a TypeScript RAG pipeline with the LlamaIndex.TS SDK.
Pinecone
Pinecone is a managed vector database built for scale. You get sub-100ms queries over billions of vectors without operating any infra.
pgvector
pgvector turns Postgres into a vector database. If you already run Postgres, this is the lowest-friction RAG store you can pick.
Build AI agents
AgentsClaude Code
Claude Code is Anthropic's terminal-native coding agent. You can extend it with custom skills, hooks, and MCP servers to build domain-specific agents that own real workflows.
LangGraph
LangGraph models agents as state machines. Instead of free-form ReAct loops, you define explicit nodes and edges, which makes complex flows debuggable and resumable.
CrewAI
CrewAI lets you compose multi-agent teams with role-based prompting. Each agent has a goal, a backstory, and tools - then a Crew orchestrates them.
OpenAI Agents SDK
The OpenAI Agents SDK is a minimal, opinionated framework for building agents on top of GPT models. It bundles tool calling, handoffs between agents, and tracing.
Build an MCP server
MCPTypeScript
Model Context Protocol (MCP) lets you expose tools, resources, and prompts to any compatible client (Claude Code, Cursor, Claude Desktop). This guide builds a TypeScript MCP server.
Python
FastMCP is the easiest way to build an MCP server in Python. You decorate functions and the framework handles schemas, transport, and routing.
Fine-tune a language model
Fine-tuningLoRA
LoRA (Low-Rank Adaptation) fine-tunes a model by training a tiny adapter instead of the full weight matrix. You get 90% of the quality at a fraction of the compute.
Unsloth
Unsloth is a drop-in optimization for LoRA fine-tuning that delivers 2-5x speedups and 60% less VRAM. Same code, faster runs.
MLX
MLX is Apple's array framework, optimized for Apple Silicon. mlx-lm fine-tunes LLMs on M-series Macs with unified memory.
Deploy an AI app
DeploymentCoolify
Coolify is a self-hosted Heroku alternative you run on a single VPS. Push to GitHub, Coolify builds and deploys.
Vercel
Vercel is the fastest way to ship a Next.js AI app: connect a repo, push to main, get a URL with edge caching, preview deploys, and built-in observability.
Railway
Railway is the sweet spot between Vercel's polish and Coolify's flexibility. You get hosted Postgres, Redis, and any Docker app from one dashboard.
Add voice to your AI app
VoiceGet structured output from LLMs
Structured OutputOpenAI
OpenAI's Structured Outputs feature guarantees your responses match a JSON schema. No more parsing failures or missing fields - the model output conforms to your schema or the request fails.
Anthropic
Anthropic's Claude models support structured output through tool use with forced tool calls. Define a schema as a tool, force Claude to use it, and extract the structured data from the tool call arguments.
Instructor
Instructor is the most popular library for structured LLM outputs. It patches the OpenAI and Anthropic SDKs to add automatic validation, retries, and streaming of Pydantic or Zod models.
Not sure where to start?
If you are new to AI development, start with Path 4. If you already know your way around, jump into the path that matches your goal.
Get Smarter About AI Dev
New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.