Code with AI

Claude Code, agents, skills and MCP explained without the hype, so you can write software with AI without breaking it.

Watch on TikTok (in Spanish) Watch on YouTube (in Spanish)

Developer in front of a monitor with three terminals where AI agents write code in parallel

How do I use AI to code without it breaking my project?

That is the question this series answers. Each video comes with a fact sheet: what it is, the key fact, how to start and my verdict. How the fact sheet works.

Before you start.

Who it is for

For anyone who already writes code, or wants to start, and uses agents like Claude Code or Codex.

What to check

Work on a branch, review every change before accepting it and never paste keys or passwords. See the full checklist.

The 8 repos that have already appeared in this series.

Each one opens its project on GitHub. The license label comes from GitHub and each repo’s LICENSE file; each repo’s warning is still in its video (in Spanish).

HyperFrames Turns HTML into video. GitHub ↗ (opens GitHub in a new tab)
License: Apache-2.0
Spec Kit Spec-driven development with AI agents. GitHub ↗ (opens GitHub in a new tab)
License: MIT
gstack Skills and workflows for working with Claude Code. GitHub ↗ (opens GitHub in a new tab)
License: MIT
Agent-Reach A command line that gives your AI agent access to Twitter/X, Reddit, YouTube, GitHub, Bilibili and more. GitHub ↗ (opens GitHub in a new tab)
License: MIT
claude-mem Memory across sessions for Claude Code. GitHub ↗ (opens GitHub in a new tab)
License: Apache-2.0
Browser Use AI agents that drive the browser. GitHub ↗ (opens GitHub in a new tab)
License: MIT
Skyvern Automate browser tasks with AI. GitHub ↗ (opens GitHub in a new tab)
License: AGPL-3.0
UI-TARS Desktop AI agent that controls your computer. GitHub ↗ (opens GitHub in a new tab)
License: Apache-2.0

Search in the directory