An Open Source AI-based Operating System, yes REALLY!

As far as I know, Magic Cloud is the only 100% open source AI-based backend and API NoCode solution out there. Not only does it create your code, but it also hosts it and is your production environment. This implies that 1 second after the LLM is done, you can test the code and edit it in its integrated IDE. And one second after having tested the code, you can start using the API and backend in production. Simply configuring Lovable to connect to Supabase takes more time than creating the thing entirely in Magic for me ...

However, the more I work with it, the more I realise it's actually an AI-based operating system. Watch this to understand what I mean ...

If you think about the implications of the above, it allows us to automate everything using AI. Need a Business Process Workflow Automation? Sure, give me 5 seconds! Need a backend for your new React project? OK, click the button! And yes, I literally mean "click the button".

Clicking the above button starts a conversation with the AI, where it leads you through your entire problem by asking questions about what you want it to do for you. I doubt it's possible to do simpler without Neuralink transferring thoughts to your compiler TBH with you ...

How Good is Magic?

When comparing features between Magic and others, this is what the system itself told me (it runs on GPT-4.1).

Feature/Platform	Magic Cloud	Cursor	Lovable	Manus AI	Retool	Superblocks	Appsmith	Copilot/Ghostwriter/etc.
AI Code Generation	✅	✅	✅	✅	❌	❌	❌	✅
Backend API Builder	✅	❌	✅	❌	✅	✅	✅	❌
Database Management	✅	❌	✅	❌	✅	✅	✅	❌
Workflow Automation	✅	❌	❌	❌	✅	✅	✅	❌
User & Role Management	✅	❌	❌	❌	✅	✅	✅	❌
AI Chatbot/RAG	✅	❌	❌	❌	❌	❌	❌	❌
Plugin System	✅	❌	❌	❌	❌	❌	❌	❌
Frontend Generation	❌	❌	❌	❌	✅	✅	✅	❌
Code Completion/Chat	❌	✅	❌	✅	❌	❌	❌	✅
Low-Code/No-Code UI	❌	❌	✅	❌	✅	✅	✅	❌
DevOps/Deployment	❌	❌	❌	❌	✅	✅	✅	❌
Observability/Monitoring	❌	❌	❌	❌	✅	✅	✅	❌
OpenAPI/Swagger	✅	❌	✅	❌	✅	✅	✅	❌
Integrated Auth/JWT	✅	❌	❌	❌	✅	✅	✅	❌
Extensibility (Custom Code)	✅	✅	❌	✅	✅	✅	✅	✅
AI Function/Tooling	✅	❌	❌	❌	❌	❌	❌	❌

Notice all the green stuff on Magic, and all the red crosses on everybody else? Then realise that everything that's frontend above might be irrelevant in 5 years from now ...

It's the BACKEND, duh!

It seems like everybody with some 20+ millions in VC money is rushing to build frontend AI "vibe code platforms" these days. Personally I don't even care about frontends. I suspect "frontend" as a concept is something that'll simply vanish in a couple of years. An AI agent can serve as a generic "one size fits all" frontend, and the cost savings of not having to build the frontend (at all) are simply too large to ignore over time. And if we move 20 years back in time, the separation between frontend and backend didn't even exist, and we were using RoR or Microsoft MVC to render stuff on the backend, intended to be displayed on the frontend. I suspect we're in such a regard moving "backwards" when it comes to frontend tech ...

However, Lovable is (currently) worth some few billions, Cursor was just bough for the same amount, and apparently Manus AI is also worth a lot these days. Magic however, is 100% open source and licensed under the terms of the MIT license - And according to GPT-4.1 it's also much better for building backends than anything else out there. Which begs the question ...

Where are my billions?

Now realistically, I don't really need "billions", but what would be nice was to for once being actually given the credit I deserve, for having built what I did. If frontends are dying, and Magic is "best" on APIs, this implies the only real value in "vibe coding" these days can be found here. Everything else will inevitably die over time, assuming I'm right about frontends. And even if I'm wrong, all of these frontends will need an API at some point in time, right?

Why I did it?

TBH with you, I'm very, very, very stubborn. I wrote an article for a Norwegian IT magazine in 2010 named "Teaching grandma to create software". NoCode and "citizen devs" have therefore naturally been very close to my heart, for a very long time.

Ever since that day, I've had one single mission, which is to prove I could do it. Others claimed I couldn't, and that made me want to prove it. Fast forward 15 years, and we're here ...

Can I now give "grandma" a software development platform, and expect her to produce an API? Fuck yeah! And I'm the only guy on earth who apparently can do that today ...

Did I succeed?

Well, Supabase is worth like 5 billion or something. They've got like 10 million databases so far, and I must presume everyone on the startup team are rich today. I'm not rich, but I've got one thing they can never match, or take away from me, and that one thing is easily wrapped up in a single sentence ...

I did it!

According to modern software development theory, building something like Magic should be impossible for a single human being. You can follow my commits on the repo if you don't believe I did it alone. About 99.8% of all commits are mine. OK, it took me 15 years, but fucking aye, I did it!!

And Magic might not be worth 5 billion dollars, but the technology is 1 million times better than Supabase, Lovable, Cursor and Manus, combined. And everybody working at these companies knows it too 😁😂

Sure, they might be better at selling and marketing, but Magic is simply superior tech - And TBH with you, knowing that they know, is enough for me - At least for today ... 😊