Notes on working with AI
The GitHub list that will make you rich (it won't)
A list of GitHub repositories promised to make money on their own. A closer look at what two of the five can actually do for a small business, and why the other three need an engineer first.
The forward every family group gets
Someone forwarded it to the family group this week: "5 AI repos that can actually make you money." A wall of GitHub links, a promise that cloning one hands you a side business, and a comment section full of people asking for the link.
I opened all five. None of them make anyone money on their own, and that's worth understanding rather than just dismissing, because two of them are genuinely useful, just not in the way the caption sold them.
What a repository actually is
An open source repository is a set of files someone else wrote and decided to share. Downloading it gets you the code, not a finished product tailored to your business, your customers, or your data. Someone still has to read it, understand what it does, adapt it, and wire it into whatever you're actually trying to build. For a business with an engineer on staff, that's a real head start. For a business without one, it's homework with no due date, and the caption never mentions the homework.
The shop owner who needs ten videos by Friday
Here's where it gets concrete. Picture a small shop running a weekend sale, same product, same offer, but they want ten versions of the ad, one per headline, to test which one actually pulls. Doing that by hand means opening a video editor ten times and re-cutting the same thirty seconds with a different line of text.
One tool from that list, Remotion, turns that problem into code. Instead of a timeline you drag clips along, you write a template that decides where the text sits, when it changes, and how it moves. Feed the template ten headlines and it renders ten videos. That's a real production tool, not a toy, and it genuinely solves the shop owner's Friday problem, but it's a developer's tool. Someone on the team has to be comfortable building and maintaining the template in the first place. Handed to a shop owner with no one who can write that template, the repository just sits there.
The one you can actually open this week
The other tool worth naming is different in kind. It's a set of instructions, built for Claude, for marketing writing: ad copy, landing pages, and structuring content so it shows up when someone asks an AI assistant about your category instead of typing a search query. There's no template to build and no code to maintain. A small business can load it into Claude and start drafting against it the same afternoon.
That's the real dividing line in a list like this, and it has nothing to do with how advanced something sounds or what it costs. It's whether an engineer needs to stand between the repository and your business, or whether you can pick it up yourself.
The other three names
The rest of that list, an open source coding agent, a tool for generating an AI avatar from a photo and a short video, and a tool for generating long-form written fiction, are all real, active projects on GitHub. I haven't tested any of the three myself, so I'm not in a position to vouch for how well they perform. What I can say is that none of them is close to something a business owner opens on a Tuesday and gets value from without engineering work first. Worth knowing they exist. Not worth clearing a week for yet.
Where the money actually comes from
The pitch behind these lists skips the part where the money gets made. It comes from the work of turning code into something a customer will pay for: design decisions, testing cycles, and the unglamorous plumbing that connects a demo to a system that runs reliably every day. None of the repositories on that list does that work for you. What they do is save you from writing certain pieces of code from scratch, which is genuinely useful, and a much smaller claim than the one usually attached to them.
Next time a list like that lands in a group chat, skip the question of which one gets you rich. Ask which one solves a problem you already have, with a level of engineering effort you can actually afford this week. For most small businesses, that shortlist is short. This week, it's two names long.