AI Development
SDKs, agentic workflows, prompt engineering, and tools for building AI-powered software.
66 topics
What this section covers
AI coding is a real engineering discipline now, not a demo. This section covers the tools, patterns, and mental models for building software with AI in the loop — whether you're pair programming, running long-horizon agents, or shipping agentic features inside your own product. You'll find deep references on coding harnesses like Claude Code and Cursor, agent SDKs including the Vercel AI SDK, and the practice-oriented topics that make any of this actually work: context engineering, agent memory systems, agentic coding workflows, and budget control for agents running unattended. It's for developers who have moved past "ask ChatGPT for a snippet" and are now designing feedback loops, harness hooks, orchestration strategies, and prompts that stay useful across a full codebase. Pair this section with AI Models when choosing the underlying LLM, and with Backend & Data when wiring an agent into persistent systems.
Add Persistent Memory to Any AI App with Mem0
Mem0 is the fastest way to give your AI app persistent, personalized memory. Here's how it works, how to integrate it in Next.js and Python, and when to use it.
Agentic Coding
Agentic coding is a development paradigm where AI agents autonomously write, test, debug, and iterate on code with minimal human intervention. Instead of act...
AI Agent Evaluation
Evaluating agents is harder than evaluating single-turn LLM outputs. An agent chains reasoning, tool calls, and intermediate decisions across many steps. A g...
AI Agent Memory Systems for Long-Running Projects
AI agents forget everything between sessions. Each time an agent wakes up, it starts with its instructions and the current task — nothing about yesterday's d...
AI Agent Orchestration
Agent orchestration is the practice of coordinating multiple AI agents to complete complex tasks. Instead of one monolithic agent doing everything, you decom...
AI Agents
An **AI agent** is a harness empowered with a role, a mission, and a scope — then pointed at work. The harness supplies the runtime (model invocation, tool e...
AI Agents vs Harnesses: The 5-Concept Stack
The words "agent" and "harness" get used interchangeably in most industry writing — people call Claude Code an agent, Cursor an agent, their test suite a har...
AI Memory Architectures: A Decision Framework
A decision framework for indie devs choosing how to add memory to their AI app in 2026 — with real pricing, latency benchmarks, and three worked examples.
AI Pair Programming
AI pair programming is a development practice where a human developer works alongside an AI assistant in real-time, combining human creativity and judgment w...
AI Playground
An AI playground is an interactive web interface where you can experiment with AI models without writing code. Providers like OpenAI, Anthropic, Google, and ...
AI SDK (Vercel)
The AI SDK is Vercel's open-source toolkit for building AI-powered applications in JavaScript and TypeScript. It provides a unified API for working with larg...
AI SDK Core
AI SDK Core is the server-side foundation of Vercel's AI SDK. It provides model-agnostic functions for text generation, structured data output, tool calling,...
AI SDK Providers
AI SDK providers are adapter packages that connect Vercel's AI SDK to specific AI model services. Each provider translates the AI SDK's unified API into the ...
AI SDK UI
The AI SDK UI library (`@ai-sdk/react`) provides React hooks for building interactive AI chat interfaces, assistants, and completion UIs. It handles the comp...
Augment Code Remote Agents: Strong Cloud Debugging Control for Technical Teams
Augment Code Remote Agents take a different approach from most cloud coding tools. Instead of hiding infrastructure details, Augment exposes more operational...
Autonomous Companies
An autonomous company is a business where AI agents handle the majority of operational work — coding, customer support, marketing, data analysis, and even st...
Build AI Memory with Your Existing Database
Skip the dedicated vector database. Build production AI memory with pgvector on Supabase, Convex vector search, or self-hosted Postgres — with working code for all three stacks.
Building Full-Stack Apps with Claude Code
Claude Code can build complete applications — frontend, backend, database, and deployment — through iterative natural language conversation in your terminal....
Building Harnesses for AI Coding Agents
A harness is not a test suite. It's not a linter. It's the runtime that wires the **Model**, **Tools**, and **Context** primitives into an autonomous loop, a...
Claude Code
Claude Code is Anthropic's agentic coding **harness** — a command-line runtime that wires a Claude model, a curated tool set, and a managed context window in...
Claude Code Hooks
Claude Code hooks are automated actions that run at specific points during a Claude Code session — before a tool runs, after a file is written, when a sessio...
Claude Code Plugins
Claude Code **plugins** are packaged, shareable directories that extend [Claude Code](./claude-code) with custom slash commands, skills, subagents, hooks, an...
Claude Code vs Cursor
Claude Code and Cursor are the two dominant AI coding **harnesses** in 2026, but they solve different problems. Claude Code is a terminal-native harness that...
Claude Dispatch: Asynchronous Task Management from Anywhere
Claude Dispatch is a feature inside Claude CoWork that lets you send tasks to your desktop AI session from your phone and return to finished work later. Prev...
Claude Managed Agents
Claude Managed Agents is Anthropic's system for orchestrating multiple AI agents from a parent agent — letting a single Claude instance spawn, direct, and sy...
Claude Vision: Image Analysis and Document Processing
Claude's vision capabilities let you send images alongside text prompts and get back structured, reasoned analysis. Every current Claude model — from Haiku t...
Cloud Coding Agents: The Practical 2026 Comparison for Solo Founders
Cloud coding agents are no longer a novelty. They are now a real production layer: you hand over a task, close your laptop, and come back to a pull request.
Coding Harnesses
A coding harness is the runtime scaffolding that wires a model, a set of tools, and a managed context window into an autonomous loop that can write, edit, an...
Context Engineering
Context engineering is the practice of designing and managing the full information environment provided to an AI model — not just the prompt, but everything ...
Cursor
Cursor is an AI-first code editor built as a fork of VS Code — it looks and works like VS Code (same extensions, themes, keybindings) but rebuilds the editin...
Cursor Cloud Agents: The Most Complete Cloud Agent Product for Fast-Moving Builders
Cursor Cloud Agents are no longer just an extension of the IDE. They are a full cloud execution layer with multiple entry points, broad model support, high p...
Designing Agent Instructions That Actually Work
Writing instructions for AI coding agents is deceptively hard. A vague system prompt produces an agent that wanders. An overly rigid one breaks when anything...
Devin AI: The Most Autonomous Cloud Coding Agent, and the Most Expensive to Scale
Devin remains the benchmark for "delegate a full software task lifecycle to an agent." It is built for end-to-end ownership: planning, coding, testing, debug...
Exa
Exa is an AI-native search engine designed for developers. Unlike traditional search engines that return links, Exa returns clean, structured content from th...
GitHub Copilot
GitHub Copilot is Microsoft's AI coding assistant — the most widely adopted AI pair programmer, with 4.7 million paid subscribers and deployment at roughly 9...
GitHub Copilot Cloud Agent: The Best Value If Your Workflow Already Lives in GitHub
GitHub Copilot Cloud Agent is not "Copilot autocomplete in the browser." It is a delegated cloud agent that takes assignments, works in a GitHub-native envir...
Giving Claude Code Browser Access
Giving Claude Code browser access unlocks a new tier of automation: the agent can open tabs, click buttons, fill forms, read page content, inspect the DOM, m...
Google ADK (Agent Development Kit)
Google ADK is an open-source framework for building, deploying, and orchestrating AI agents. It's designed to make agent development feel like traditional so...
Google Generative AI SDK
The Google Generative AI SDK (`@google/generative-ai`) is Google's official JavaScript/TypeScript library for interacting with Gemini models. It provides dir...
Google Jules: High Concurrency and Simple Task Economics, With One Big Access Constraint
Google Jules is one of the most interesting cloud coding agents for solo founders because it combines a generous free tier with clean task-based limits. You ...
Human-in-the-Loop
Human-in-the-loop (HITL) is a design pattern where AI systems operate autonomously but pause at critical decision points for human review and approval. In ag...
Managing AI Agent Budgets and Cost Control
Running AI agents costs money. Every prompt, every tool call, every reasoning step consumes tokens that translate directly to dollars. When you have a single...
MCP (Model Context Protocol)
Model Context Protocol (MCP) is an open standard created by Anthropic that enables AI models to securely connect to external data sources and tools. It provi...
Multi-Agent Task Delegation: Patterns and Anti-Patterns
When you have multiple AI agents working together — a CEO agent, a CTO, engineers, a researcher, a marketer — the hardest problem is not getting any single a...
Object Generation
Object generation (also called structured output) is the process of using AI models to produce typed, structured data instead of free-form text. Rather than ...
Official MCP Servers by Vendor
Model Context Protocol (MCP) lets AI coding tools connect directly to the services you build with. Instead of describing your Cloudflare setup in a CLAUDE.md...
Official Skills for Your AI Coding Stack
The major infrastructure vendors are publishing their own Claude Code skills, Cursor rules, and AI coding context files. Instead of writing `.mcp.json`, `CLA...
OpenAI Assistants API
The OpenAI Assistants API lets you build AI assistants that can use tools like code interpreter, file search, and custom functions. Unlike the Chat Completio...
OpenAI Codex Cloud: When the Cheapest Entry Point Is Also Good Enough
OpenAI Codex Cloud is one of the easiest ways to start using asynchronous cloud coding agents without committing to a high monthly bill. The $8/month Go plan...
OpenAI Moderation API
The OpenAI Moderation API is a free tool that checks whether text content violates OpenAI's usage policies. It classifies content across categories like hate...
OpenAI Responses API
The Responses API is OpenAI's latest API for building AI applications. It combines the capabilities of the Chat Completions API with built-in tools like web ...
OpenAI Text-to-Speech (TTS)
OpenAI's Text-to-Speech API converts text into natural-sounding spoken audio. It offers multiple voice options, two quality tiers, and supports various outpu...
Prompt Engineering
Prompt engineering is the process of crafting inputs to AI systems to get desired outputs, an essential skill for working with generative AI tools.
Prompt Engineering for Claude: Advanced Techniques
Prompt engineering for Claude is about structuring your inputs to consistently get the outputs you need. Claude's architecture responds well to specific patt...
Prompt-Driven Development
Prompt-driven development (PDD) is a software methodology where natural language prompts replace manual coding as the primary means of building software. Ins...
Ralph Loops (Rapid Agentic Loop for Programmatic Hacking)
A Ralph loop is a tight, iterative cycle where an AI agent repeatedly writes code, runs validation, observes results, and corrects itself until the task is c...
Tavily
Tavily is a search API built specifically for AI applications and RAG (Retrieval-Augmented Generation) pipelines. It returns clean, structured search results...
Team Collaboration with Claude Code CoWork
Claude CoWork brings Claude Code's agentic capabilities beyond the terminal and into everyday knowledge work. Launched in January 2026 as a research preview,...
Temporal Memory with Zep — When Your App Needs to Track Change
Zep and the open-source Graphiti engine give your AI app bi-temporal memory: track when facts became true and when they stopped. Deep dive on architecture, pricing, code, and when to choose Zep over Mem0 or pgvector.
The Harness Orchestration Loop
The harness doesn't just validate code. It runs a loop: observe the current state, plan the next action, act, verify the result, and repeat. This observe → p...
Unstructured
Unstructured is a data processing library and API for extracting content from documents like PDFs, Word files, HTML pages, images, and more. It converts unst...
Upstage
Upstage is a South Korean AI company that provides document AI and language models. Their flagship product, Document AI, specializes in extracting and unders...
Vibe Coding
Vibe coding is a software development approach where the developer describes what they want in natural language and lets AI generate the implementation. Coin...
Why Your AI App Forgets Everything
LLM APIs are stateless by design — every session starts from zero. Here's why that kills retention, and the four architectural approaches for fixing it.
Windsurf (Cascade): The Compliance-Strong Cloud Agent Option With Low Entry Cost
Windsurf (Cascade) has carved out a distinct position in the cloud coding agent market: affordable entry pricing, strong enterprise compliance posture, and b...
Working with the Claude API: Authentication, Rate Limits, and Scaling
The Claude API gives you programmatic access to Claude's capabilities — text generation, vision, tool use, and more — for building production applications. T...