Goodbye, GitHub

When I was a kid, Nintendo was so dominant in the video game space that I remember getting a Sony PlayStation and rolling my eyes when my mother called it "the Nintendo". For some people, the brand Nintendo had replaced the phrase "video game".

The same thing has happened with "Git" and "GitHub". They aren't the same thing.

Git is version control software. It was originally created by Linus Torvalds, the Linux guy, to track changes to code files and let distributed teams collaborate effectively without stepping on each other. It's open source, it's free, and you can install and run it anywhere.

GitHub is a company that built a hosted service on top of Git. They have a free plan, a web interface, and additional features for collaboration and automation bolted on. The free tier is fine for individuals, but most of the really powerful additional features are behind the paywall.

Is GitHub useful? Yes, but you also need to be aware that they're selling you something that is inherently free. But convenience is sometimes worth paying for, and the brand is so dominant that when beginners think it is the only way, they don't take the time to consider other options.

This is a post about those other options and how/why I've moved Skill Foundry off of GitHub.

All Skill Foundry Curriculum is in Git

Because I teach software development, I use Git the obvious way: sharing exercises and example code with Skill Foundry learners.

But what people don't realize is that I run almost all of the business's content on it, too. All of the lessons, interactives, knowledge checks, blog posts, and more live in Git. Even non-technical team members who help me with things like editing and instructional design must learn Git.

So where the content lives, who has access, and the controls for versioning, backup, and recovery are not trivial questions I had to answer. For years, the answer was GitHub. Then, it stopped being a good answer. One reason was Microsoft.

Microsoft Bought GitHub

To say that my early career was all-in on the Microsoft stack is not an exaggeration. Many enterprises cluster around specific stacks because integration is usually easier. Back in the day, the official C# and .NET Framework were also Windows-only, so that was the stack I was comfortable with. So when Microsoft bought GitHub, my response was to shrug.

But then, the enshittification started. The product got heavier, reliability started to decline, and then AI came along. Microsoft, much like most of the big tech firms, has started cramming AI into every layer of its stack, including GitHub, and reliability has dropped even further. Some of it is on Microsoft; some of it is on the explosion of vibe coding, where autonomous agents are flooding the system. This is already leading to some cuts to once-free services and tighter limits, and I expect it to get much worse.

As someone who creates content, I also have a bigger concern. The frontier labs trained on the world's copyrighted materials without asking anyone first. The copyright holders are pushing back, and there are now dozens of lawsuits working through the courts, and they are not all going the labs' way.

So, when Microsoft announces that the default is now allowing them to train their models on our content, you'll have to forgive me if I don't trust them. I'm not saying that the opt-out is definitely a lie; it's just that I've seen how much these companies' promises are worth when the financial incentives point the other way, and I'm not putting my business and content on the honor system.

Another thing worth mentioning is that even if these companies get caught violating policies and privacy, the legal response is usually a slap-on-the-wrist fine, which isn't a disincentive for companies worth trillions. So I'm skeptical that even if caught, I would ever be appropriately compensated.

The Mentorship Offering

A second reason that I decided to move off of GitHub was for my mentorship plan members. While the core Skill Foundry subscription is self-paced, the mentorship plan includes personal mentorship from professional developers, including me.

For the mentorship offering, I wanted each learner to have their own private, controlled Git repository, shared with me and my team, so that we can provide detailed feedback on their work.

Now, GitHub can technically do this, but it would mean putting my learners' work on someone else's server, under someone else's terms, and subject to the same privacy issues I raised earlier. Also, and more importantly, I can't guarantee that GitHub will remain free. I do everything I can to ensure the tools I use in the courseware are free, and hosting some of them on my own servers helps me shield my learners from those costs.

I wanted more control, so I put the time in to get it.

Moving to Self-Hosting

Here's what I did at a high level:

1. I rented a small VPS: The cloud is just computers sitting in data centers. You've probably heard of AWS and Azure, but there are many providers that let you rent Virtual Private Servers that are slices of bigger machines that you fully control. I rented mine from Hetzner, with Ubuntu Linux, in one of their EU data centers, because the EU has much stronger privacy regulations.
   
   The cost of that server is about $10 / month, and it can do things that would cost hundreds of dollars per month on GitHub.
2. Installed Docker and Caddy: Docker packages software to make it easy to deploy to servers. Caddy is software that handles web traffic and HTTPS for you.
3. Wrote a Docker Compose File for Forgejo: Forgejo is an open-source, self-hosted Git service with a full web interface and basic automations. My learners get a web interface similar to GitHub, but I own the server it runs on.
4. Added a DNS Record: Skill Foundry now has a subdomain pointing to the Forgejo application, so anyone can reach it.
5. Migrated from GitHub: Forgejo also has an import tool that lets you quickly move repositories from GitHub into Foregjo.
6. Testing Period: The team and I used Forgejo internally for a few weeks, making sure everything pulled over and that it was all working well.
7. Deletion: I logged into GitHub, and started deleting. At this point, almost everything is moved over.

The initial setup and migration took less than three hours, because I know my way around Linux and Docker.

The Tradeoffs

Self-hosting isn't for everyone, but it is a skill that I think most IT professionals should have. Not to pick on developers, but I'm going to pick on developers. Many professional developers have no idea how security, networks, file storage, and resource allocation really work. For the vibe coding crowd, it's even worse, which is why you see memes like this floating around:

If you don't know why that's funny, it's ok; I'll explain the joke. Localhost refers to your computer, so giving someone links to localhost would try to open it on their computer, where the application the poster is trying to show off doesn't live, so the links wouldn't work.

Some of the upsides of self-hosting are privacy, control, and cost savings. But the downside is that you're now responsible for the server and your applications. Nobody is automatically handling backups; if the server goes down and you haven't configured alerts, you won't know. If bad things happen at 3am on a Sunday, you have to get up and fix them.

But, self-hosting has gotten a lot easier over the years. Backup jobs can be scripted. Software can be installed to monitor and alert you when things go wrong, and managing applications with Docker Compose is genuinely easy if you understand how things work.

Also, running some of my own infrastructure instead of renting it from a SaaS provider literally saves me over $10,000 per year in business expenses.

Upcoming Self-Hosting Course

Trust me when I say this: SaaS is not going to get any cheaper. As providers stuff expensive AI into their products and hardware costs continue to rise due to AI data center demand, the prices for these services will go up.

If you become an independent contractor or work for a small or medium business, you can save a lot of money by putting the applications that make economic sense onto your own servers.

And so, later this month, I will be launching the newest Skill Foundry course on self-hosting fundamentals.

It won't make you a certified Linux administrator, and it doesn't pretend to. It walks you through the actual fundamentals: acquiring a domain, configuring DNS, standing up a VPS, basic server hardening, installing a reverse proxy, hosting a static site, installing and configuring Docker and Docker Compose, and serving a real containerized application. In the course, I use WordPress as the example, but the workflow is the same for any containerized software.

For my learners who are job seekers: this skill will let you build far more impressive portfolio sites than what you can do with GitHub Pages. A live app you deployed and maintain yourself, on your own domain, says far more about your fundamental skills than another to-do app on someone else's free tier. And most importantly, it will make you a better developer, because when you understand the infrastructure your applications run on, you will stop thinking of it as somebody else's magic.

Dependency, Privacy, and Ownership

Renting a capability is a dependency. Dependencies are fine right up until the vendor decides to change the terms by adding bloat, forcing in AI, raising prices, or changing what they're allowed to do with your data. If you aren't at least capable of self-hosting, you have no recourse.

I consider self-hosting whenever I'm evaluating solutions for my home and business, and I'm choosing the self-hosting path more often these days for two reasons:

1. SaaS privacy keeps eroding, because the incentive to monetize customer data is compounding every year.
2. Inflation keeps pushing subscription prices up, and with the jumps in hardware costs due to AI data centers, that cost will be passed on to you.

Everywhere you leak, the world hangs a bucket. And being able to run your own infrastructure means there is a limit to what you have to tolerate from vendors.

And if you are optimistic about the future of AI, hosting local, open-source models is the most likely path forward. You'd best get learning!