Happy Monday,
I paid fourteen dollars for an iced coffee this morning.
Eight minutes to make. Order still wrong.
Meanwhile a guy at a car meet last weekend was pouring the RIGHT drink in thirty seconds out of a converted Home Depot cart he built for under a grand.
One of them is running a business. The other one has a vibe. We have an Idea.
In Todays Edition, TL;DR
the home depot cart that out earned a six figure food truck
how to clone a competitor's ad with AI in under an hour
the scheduling layer every AI agent stack is missing
Selling on livestream may be the future…
The Monday Idea
💡 A portable nitro iced coffee cart with low startup costs, high profit margins, and no espresso machine, no lease, no employees.
The Problem: The coffee industry has convinced itself that complexity equals quality. Espresso machines the price of a car down payment. Baristas measuring grounds by the gram. twenty minutes per cup. absolutely ridiculous.
95% of orders have cream, ice, and syrup in them. At that point the actual coffee is irrelevant.
The thousand dollar machine isn't making better coffee. It's making the owner feel better about the coffee AND the outdoor event market is embarrassingly large. Car shows, air shows, farmers markets, festivals. Tens of thousands of events every weekend. Most have one or zero cold coffee options.
🛠 Solution: Pre brewed nitro coffee in a bag in box container, tapped through a converted Home Depot tool cart, served in 30 seconds. Cold brew on tap, minus the real estate and the equipment debt. Shelf stable for three to six months. Costs roughly a dollar fifty per serving to make, sells for eight to nine at events. No grinding, no brewing on site, no coil dying two hours from civilization.
The cart could be a gutted Husky tool chest. Tap system, nitrogen tank, a sign. Total build: under a thousand. Menu: two or three options. The simplicity is the product.
The Business Model:
Owner operator: You work the events yourself, take the full margin on every cup. This is the validation phase. Low overhead, high learning.
Operator model: You build the cart, source the product, find the locations. A second person runs it. Revenue splits roughly 70/30. They handle labor, you handle logistics.
Fleet: Multiple carts, multiple operators, multiple events running simultaneously. You become the supply chain and the location network. Each cart is a revenue unit you don't personally have to staff.
How did we do? 👇🏻
Founder’s Tutorial
how to clone a competitor's ad with AI (in under an hour)
The concept: find a UGC video that's already working for a competitor, feed it to an AI agent, and have it recreate the entire video with your product. Automatically.
Here's how it works.
What you need
Two tools. Claude Code (free desktop app) and Arcads, which is where the video gets generated using a model called CogVideoX Dance 2.0. Claude handles the analysis and the prompting. Arcads handles the generation.
Step 1: Find a winning ad
Go to TikTok Creative Center or Meta Ads Library. Both are public. Both show you what ads are running and how they're performing. Find a UGC video in your niche that's clearly working. Download it.
Step 2: Set up Claude Code
Open Claude Code, connect it to a local folder, and install the UGC skill from the repo linked below. This teaches Claude the entire workflow so you don't have to re-explain it every time. Then grab your Arcads API key from settings and paste it into Claude to connect the two tools.
Step 3: Feed it your inputs
Drag three things into Claude: the reference video, a photo of your product, and any other assets you want in the shot. Then prompt it: "Replace the product in the reference video with the provided images."
Claude will watch the video, transcribe the audio, break down the beat structure scene by scene, rewrite the script for your product, and fire the full prompt to Arcads automatically.
How do you guys like the Tutorial Section?
Napkin Note Ideas
💡CalQueue: a shared scheduling layer that AI agents query before booking anything.
The Problem: Most people running AI agents have at least three tools touching their calendar. A sales agent books discovery calls. A recruiting tool schedules interviews. A personal assistant blocks focus time. None of them know what the others are doing.
The agents aren't broken. The infrastructure is missing.
Existing calendar tools were built for humans booking with humans. There's no API layer designed for agents to check conflict logic before committing. So they don't check. They just book.
🛠 Solution: CalQueue sits between the agents and the calendar. Availability windows, meeting preferences, buffer rules, and conflict logic live in one place. Before any agent commits a booking, it queries CalQueue first.
The Business Model
Individual: Monthly flat rate, single user rules and availability windows
Teams: Per seat monthly, multi agent support and shared conflict rules
Enterprise: White labeled endpoints, SSO, audit logs, annual contract
Weekly Gems Worth Reading
Check out this guy who made a Telegram bot that makes 100k/year
GaryVee predicts selling products on livestream will be HUGE 👀
Do you think AI is the biggest opportunity for creative people?? Will Taste be the new moat?
That’s all guys
That's Monday. A coffee cart that clears more than most side hustles, an AI workflow that cuts ad production to an hour, and a scheduling problem nobody's solved for the agent era yet. Good week to find a farmers market and count cups.
Talk Friday.