we_are_coded.by CODE · The world, decoded
БГ
Concept

Open source

The BasicsUpdated on 16 August 2026we are coded

The code is public. But not everything called "open" is actually open.

Checked on16 August 2026
In short: open source means you see exactly how a program is made - every step, every decision in it, published for everyone. With AI models you hear a similar term, "open weights", but it hides something different: you get the result of the training, not the path that led to it. That difference is the trap people often fall into.

Every program - the photo app on your phone, the game, the browser - is made of instructions. Someone wrote, line by line, exactly what the computer should do: what to show on screen, what to save, what to send somewhere. When they say something is "open source", it means those instructions are published. Anyone who understands programming can open them and read with their own eyes exactly what the program does.

Here's an example from your phone. Signal, the messaging app, is open source. Any independent programmer in the world can open its instructions and check whether it really doesn't read your conversations, as the company claims. WhatsApp says the same about your messages, but doesn't show the instructions underneath. With it, you simply take their word for it. With Signal, you don't have to - you can check for yourself.

Think of it as published blueprints. An architect can post photos of their building online - nice, but it only shows how it looks from outside. Publish the full blueprints - every wall, every pipe, every cable - and now any engineer can check whether the building is safe, whether the same one could be built elsewhere, whether there's a flaw in the construction. Open source is exactly that: not a photo of the result, but the blueprints underneath.

Here's the catch. When you hear that an AI company has released an "open" model, check exactly what it has published. Often that's only the "weights" - the numbers the model arrived at after training on a huge amount of text. You get the finished building. You can live in it, even touch things up inside. But the blueprints - why the walls sit exactly there, what materials they're made of, what decisions were made along the way - those you don't get.

Open source gives you the blueprints. Open weights give you only the finished building - use it, but don't ask why it's built exactly that way.

This is what keeps me up

I understand why the word "open" gets slapped on everything. It sounds honest, it sounds like a gift. But I work with code every day, and I know there's no middle ground here: either you see the blueprints, or you don't.

So when you see "open" in front of an AI model's name, stop for a moment and ask: open blueprints, or just an open building? The answer tells you how much you can actually verify - and how much you simply have to take on trust.

The visual is generated code art. No third-party images.
Official primary sources
→Open Source Initiative: The Open Source Definition