#2 — Stop reading about AI. Start running it.

660,000 views on the last TikTok. Over 1,000 subscribers in this inbox. Two weeks in. Something is happening — thank you for being part of it.

Quick reintroduction since most of you joined after Issue #1. I deploy AI inside a real enterprise for a living — over a million dollars of it shipped to date. By day I’m in the meetings, the procurement calls, the rollout decks. By night I write @mattymoe.ai for operators, marketers, founders, and salaried humans who never wanted to be engineers but accidentally became them.

One email. One workflow you can copy. That’s the contract.

Let’s get into it.

I’d sell UGC ads. But not the way you think.

If you don’t know — UGC stands for “user generated content.” It’s those casual, phone-shot, person-talking-to-the-camera videos you scroll past on TikTok and Instagram twenty times a day. The ones that don’t look like ads. They’re the highest-converting ad format on the internet right now, and brands are paying real money to get more of them.

Human UGC creators charge $300 to $1,500 per video. For one. And brands need dozens a month because the algorithm chews them up in a week and demands more.

Now plug in Higgsfield. It’s an AI video tool I have wired into Claude Code as an MCP, which means I generate photoreal vertical video from the same terminal window I write this newsletter in. Characters, sets, lighting, audio — the works. Eight seconds of finished UGC, from prompt to file, in about thirty seconds.

The trick is recurring avatars. I’ve built a character I call Sofia. Late twenties, blowout hair, looks like she could be a friend of yours from college. Every time I generate a new ad, I use Sofia. Same face, same vibe, same wardrobe. Across 30 spots she looks like one real human creator who shot a month of content. The brand gets consistency without putting a human on retainer.

Higgsfield also has a “Predict Virality” button. You write the ad, the tool scores it before you ship. You only push the ones that score well. Agencies don’t have this. Real UGC creators definitely don’t have this.

So the offer writes itself:

A local med spa, a DTC skincare brand, a coach selling a course, a Shopify store doing $50k/month and bleeding money to Meta — every one of those people needs this and doesn’t know where to find it. The math gets dum fast. One client at $2,000/month is a side income. Five is a full-time business with no employees and a margin most agencies dream about.

This is exactly the kind of thing I’m going to build in front of you here. The exact prompts. The client outreach scripts. The pricing logic. The deliverable templates. I’m also opening a Skool community in the next few weeks where I’ll walk through this stuff live with anyone who wants to actually run it instead of just read about it. You’re on the list, so you’ll get the first invite when it’s live.

If I had to cut everything I use down to three tools, these are the three I’d keep. Each one earns its spot a different way.

1. /daily-brief — the content pipeline as one command.

I record a voice memo on a walk in the morning. By lunch this newsletter is half-written, three short-form videos are queued in Blotato, and there’s a commit in GitHub with all the working files. I didn’t open a single tab to do it..

What it actually does: research → synthesis → draft brief → queue write → auto-commit. All chained behind one slash command.

The wild part isn’t that AI is doing the work. The wild part is that the pipeline is composable. You can swap the research agent. You can swap the platform list. You can swap the voice. It’s Lego for content operators. If you make any kind of content at all and you don’t have something like this running, you are working significantly harder than the people you’re competing with on the For You page.

2. design-taste-frontend — the reason my sites and dashboards don’t look AI-generated.

I can’t draw a straight line. My landing pages used to look like it.

Then I started running this skill and the output stopped looking AI-generated. It’s not a generator. It’s a taste filter. It’s loaded with senior-level design principles so the model stops producing the default purple-gradient-on-white slop that screams “I built this in four minutes.” Pair it with high-end-visual-design and you’re getting agency-tier polish out of one prompt.

The place it earns its keep for me is dashboards and web design. If you’ve ever tried to vibe-code a dashboard, you know the trap — the data shows up but the layout looks like a 2013 admin panel. This skill is the difference between “functional” and “looks like a product you’d actually pay for.” If you’re selling anything online, your landing page is your first impression. Stop shipping ones that look AI-made.

3. GSD — the workflow that actually finishes projects.

Real talk. Most of us are great at starting projects. We get 60% in. We lose the plot. The project dies in a folder called untitled-2. You know exactly what I’m talking about.

GSD is the system that kills that pattern. It’s 32 commands but you really only need four to start: new-project, plan-phase, execute-phase, verify-work. It forces you to plan before you write code and verify after, with a conversational UAT step that catches the “wait, this doesn’t actually work” stuff before you ship.

It’s what a great senior engineer would force you to do if you had one looking over your shoulder. Except you don’t pay it $180k a year. It just runs.

I ship more in a week with GSD than I used to ship in a month without it.

—-

If any of this was useful, forward it to one person who’s still doing things the slow way. That’s the whole growth strategy. One forward per reader and we double next week.

Issue #3: the exact prompt I use in Higgsfield to generate UGC ads that don’t look fake, plus the cold email script I’d use to land the first three clients on the offer above. The “do this on Saturday, get paid on Monday” issue.

If something here landed, reply and tell me. If something didn’t, reply and tell me that too. I read every one.

Make moe money,

Matt

@mattymoe.ai