Hey there,

Lots of new joiners this week. Welcome to my weekly recap, where I talk about what I got done this week and the AI projects I've been working on.

This was the biggest thing for me this week. I made a new video about Grok Bot because they'd released a lot of new features, and it's gotten a lot better.

My honest take: it's finally fulfilling the promise of OpenClaw. I'm handing over more and more tasks, and it handles them autonomously.

And the video was liked and reposted by Elon, which was obviously a big milestone for me and made me very happy 😄

The fact that it happened in a week when Elon pulled off the most memorable product demonstration of all time makes it even cooler.

The video itself did very well on YouTube too. Here's the Grok Bot video if you want to check it out:

My goal with Grok Bot is to automate my business and eventually, once it gets good enough, automate several businesses and have them run autonomously.

I don't think this is a far-fetched future. I think we're close, and that this will be possible soon.

There will likely be a short arbitrage moment in history where some people know about this and can really capitalize on it. I want to try to be ready for that. This is what I'm working towards.

And if you haven't grabbed it yet, you’ll find the free Grok Bot Playbook under the free resources link below.

Last week I launched FrameJam, a plugin for Claude Code, ChatGPT, and Cursor to help you edit videos.

I built it for myself, and I'm using it in my editing process. It's been super helpful.

This week I've been doing more work on the product, and I introduced a Playbooks feature.

A playbook gives your agent a process for making a particular kind of video: what the agent needs to bring, which steps to follow, and what the finished video should look like (kind of like a skill). So you can pick the kind of video you want to make and give the agent a starting point.

Here's the full tutorial:

Since I've been using AI agents to edit videos a lot, I noticed the same thing happening over and over again.

If you want Claude to analyze a video, use it as a reference, or edit it, it first needs to understand what's actually in it. Basically, it needs to watch it.

Claude can't actually watch a video, so it takes screenshots and tries to understand the video from those.

I built a tool that standardizes and improves that process. It's called Claude Eyes.

It processes a video once and gives the agent sheets of labelled frames, a transcript with timestamps, text shown on screen, and details about the sound and editing style. The agent can then search for the right moment and zoom in to read small text or code.

And here's the GitHub repo if you want to try it out.

I'm planning to do an exclusive AI video editing masterclass next week inside my paid community, where I walk you through my personal process of editing videos with Claude. Videos that are generating hundreds of thousands of views per week. I used to pay a video editor to do this work for me, now I just let AI do it.

If at least 5 people register and sign up for it, I'll do it.

It'll be on Monday, October 19, at 5pm Berlin time (CEST) / 8am Pacific (PDT).

And if you just want the free stuff, that’s fine too. Here are all of my free skills and resources: Prompt Warrior Resources

My kid and partner came back from their trip on Thursday. They'd been gone for a week.

I had a really strong week of productivity because of it, but I'm happy to have the kid back, of course.

I shipped 4 videos in that time, and I want to try to keep it up.

My focus is on longform videos right now. My shortform videos are suffering a bit, but I think it's a necessary sacrifice to keep the momentum going on longform and X.

Have a great weekend everyone!

Moritz