Guides.
Long-form, opinionated reference docs for AI-era web development.
-
How to Actually Deploy Your Vibe-Coded App
Choose a host, lock down your secrets, set up a domain, monitor for blowups, and roll back without crying.
-
Stop Burning Tokens
Real math for an 8-hour coding session at every model tier. When to compact, when to cache, when to start fresh.
-
Cursor for Web Developers
Workspaces, modes, hooks, MCP, rules, background agents — the complete reference, web-dev shaped.
-
Running 4 Agents at Once
Worktrees, Conductor, and the review discipline you need to not turn parallelism into chaos.
-
OpenCode from Zero to Production
Install, configure, choose a provider, install the MCP that actually helps, and compare it honestly to Claude Code and Codex CLI.
-
AGENTS.md for Web Projects
What to actually put in
AGENTS.mdfor Next, Astro, SvelteKit, Remix — and what to leave out. With a live generator that builds the file for you. -
MCP Servers for Web Devs
The seven MCP servers that earn their keep in a 2026 web-dev stack — and the token tax each one adds to every turn.
-
Picking a Model in 2026
Decision tree for Claude Opus 4.7, Sonnet 4.6, GPT-5.3 Codex, Composer 2.5, and Gemini 3.1 Pro. With a live picker that gives you a model and a price.
-
Testing AI-Written Code
Contracts not implementations. The right ratio of unit/integration/e2e for a 2026 codebase. When to let the agent write tests, when not to.
-
Self-Hosting an AI Coding Stack
Ollama, Cline, Aider, and a local Qwen or DeepSeek model. When local actually wins, when it doesn't, and the payback math.
-
Streaming AI in Web Apps
Vercel AI SDK 5, SSE,
streamText, tool calls without footguns, and a budget estimator that tells you what the bill will actually be. -
Securing AI-Generated Code
Slopsquatting, prompt injection, leaked secrets, and over-permissioned agents — the four ways AI ships you a CVE, with a 12-control self-audit.
-
Prompting for UI Code
Type-first specs, design tokens, the states block, accessibility in one line — with a live prompt-quality scorer that grades a pasted prompt against eight criteria.
-
Building MCP Servers
Write your own MCP server in TS. stdio vs Streamable HTTP, auth and sessions, ops — with a live scaffold generator that emits a runnable starter file.
-
AI for Performance Optimization
Prompts that actually move LCP, INP, and CLS. The unnecessary-client-component hunt. A live vitals diagnostic that emits the right prompt for what's broken.
-
Observability for AI Features
OpenTelemetry GenAI conventions, the seven attributes that matter, SLOs for streaming endpoints, vendor tree, and a live error-budget calculator.
-
Claude Code for Web Developers
Permissions, MCP, subagents, session forking, hooks, and a live permission profile builder for
.claude/settings.json. -
Codex CLI for Web Developers
Goal mode, security review, MCP, and a live task router that picks Codex vs Claude Code vs Cursor for the job.
-
Accessibility for AI-Built Sites
WCAG 2.2 AA for vibe-coded UIs. Prompt contracts, keyboard/focus, native dialog, axe + Playwright, and a 12-point audit checklist.
-
CI/CD for Agent-Written Code
GitHub Actions playbook for AI-era repos. Secret gates, dependency review, preview deploys, and a stack picker that emits workflow YAML.
-
RAG for Web Apps
Chunking, embeddings, pgvector vs Pinecone vs Vectorize, AI SDK 5 retrieval routes, tenant isolation, and a monthly cost estimator.
-
Design-to-Code with AI
Stitch, Figma MCP, Impeccable, and in-repo agents. When generators help, when they hurt, and a picker for your starting input.