Free guide

The Three Videos

The only three AI videos I would send someone starting from scratch. Chat, then cowork, then code, with an honest verdict on each and what the order is for.


Use this with your AI

Copy the complete guide as clean Markdown, then paste it into ChatGPT or Claude.

Someone asked me what to watch to get up to speed on AI.

I've sat through a lot of bad AI content this year. Most of it is either a demo of something you can't use or a lecture on something you don't need yet. So here are the only three I'd actually send.

The order is chat, then cowork, then code.

1. Jeff Su, Claude Cowork in under 20 minutes

What it is. A straight walkthrough of Anthropic's desktop app and its seven capabilities: local file access, persistent memory, connectors, skills, projects, the browser extension, and scheduled tasks.

The verdict. The best pure-usage video on this list and the only one I'd call genuinely beginner friendly. No theory, no setup pain, nothing to install before you can follow along. If you're still copying and pasting between a chat window and your actual work, this is the video that ends that habit.

It's also fully timestamped, so you can jump to the capability you care about instead of watching linearly.

2. Greg Isenberg with Ras Mic, how agents and skills work

What it is. A conversation rather than a tutorial. Context windows, what a skill actually is under the hood, how to build one, and how to keep sharpening it.

The verdict. This is the one that changes your mental model. The useful claim: skills load lazily. Only the name and description sit in context until the agent decides it needs the full file, which is why a folder of thirty skills costs you almost nothing until one gets used.

His framing is that the models are already good, and the differentiator is the context and harness you build around them. That's an unusual thing to hear from someone whose audience wants model news.

3. Nate Herk, Claude Code as your executive assistant

What it is. He builds a working AI executive assistant in four phases: set up the project, load it with context and rules, add your first skills and subagents, then grow it over time.

The verdict. Watch this third. It's the most in depth of the three and by far the most rewarding, and it will half make sense if you haven't done the other two. The demo he opens with is the honest sell: tasks that ran before he woke up.

I run a version of this setup every day. It knows my clients, my priorities, my writing voice, and it drafts off all three. It's the single highest-leverage thing I've built for myself, and it started with roughly this video.

Newsletter

Get the next one

One useful email a week on AI for people running businesses. Unsubscribe anytime.