I am running a new offer to solo car detailers, I will book you $1,000 in free car detailing appointments using meta ads and auto AI texting system. This offer is working really well I am about to close a $3,000 set up tomorrow at noon for the first client. Running this offer has forced me to really break down what makes a good local mobile car detailing ad. I live in Pullman Washington (pop. 30,000) and i have been dominating this local market with my little solo car detailing business Refresh Mobile Detailing for the passed 2 years. The first car detailer who claimed this offer lives in Spokane, I kicked off his ad campaign and saw my lead cost up at $13 instead of $2 which is what I was getting in my hometown. The reason was Spokane had competition! I had to get better with my video ad creatives. I went surgical, every second of a 60 second video ad was broken into components so I could test different components. I created new offers and new packages, and watched as my lead cost started to drop. hell yeah. But it took a lot of manual work, now that I know this offer to car detailers is about to get me paid, I am going to be working with 5 - 10 car detailers at a time, which means I need a system that can help me book them thousands of dollars in jobs with low cost per lead and booked appointment. I also had a nice fellow named Leandro ask me about my AI ad creative system. So the video I attached is my response to his question but I wanted to share it with everyone. The following video breaks down how I am going about creating a volume of video ad creatives that are high quality and have the ability to Frankenstein many different components together in order to run experiments to hit key metrics. I will be building this in the following weeks... enjoy https://www.loom.com/share/f9fa9a625eac4ba8a381eff5cdab2ff5

The video you just watched made itself. I gave Claude Fable 5 one prompt inside Claude Code, walked away, and came back to a finished video I never filmed, edited, or even watched while it was being built. In this one I show you what Anthropic's new Mythos-class model can actually do, then break down exactly how it wrote the script, cloned my voice, rendered the avatar, built every motion graphic, and edited the whole thing on its own.

For a while now I've struggled to make the change from n8n to Claude Code / Hermes with my content. I make videos for hyper-specific use cases, and that was a lot easier to do with n8n. But I've been spending more time with coding agents and AI operating systems, and I'm finally getting the hang of them to the point where I understand the new format I'll be building videos around: AI "Employees" Hermes profiles, or tailored skill files + cron jobs, that each handle one hyper-specific problem. This first video is a super simple use case, but for contractors it starts to claw back the admin time that chips away at your week. In the tutorial I cover: - Getting a server live with Hermes - Setting up a Context Operating System tailored to residential builders - Connecting it to Telegram - Giving your employee access to Gmail + Google Drive Hope you enjoy. There's a lot more coming, I have a lot to say about making these coding agents solve real business problems. 👉 https://youtu.be/AqjNpddohqk

Rebuilt my Hermes setup from scratch 3 times. I broke down the full mental model, what each folder actually does, how I use SOUL.md to enforce structure, why hermes profiles are the key, and how I'm packaging all of this to sell AI employees to contractors. The big unlock for me was separating the agent's brain from the filing cabinet. Once that clicked, everything else fell into place. I recorded a loom to walk through my thoughts! At the end I talk about how I am selling these systems 😁 https://www.loom.com/share/70da16b66039490dbb9191d95e7126fe

You can't build a real company on someone else's foundation. So what's the solution? You must build with data first. Here's how it works: 1. Figure out what your customer is trying to achieve 2. Map out all the things that need to happen for them to achieve it 3. Build systems that capture data about those things 4. Use that data to predict outcomes The AI just turns messy input into clean numbers. Your data and equations make the actual decisions. This is how you build something that lasts 10 years instead of 6 months. I just released a video breaking down the full framework. It's 15 minutes. Link in comments. Stop building wrappers. Start building moats. https://youtu.be/iyZwe6kAmfQ