Artificial Intelligence as an element of power (And that's even before AGI arrives)

During my years at Boğaziçi University, I had a friend named Emre. He was studying electrical and electronics engineering. The software world of those days was more naive than it is today. There was a concept called "shareware." You would install the program, open it, and a window would pop up. It would ask you for a license code. If you entered the correct code, the program would open and work with all its features. Most of the time, you didn't even need that. Emre would open an assembler, go into the code, examine a few lines, and change a jump or two. That popup would never appear again. The program would be "licensed" forever.

Watching this back then was impressive. Looking at it today, it seems like an almost nostalgic skill. Because what Emre did is no longer a task that one person would do with patience. The picture that has emerged with Anthropic's MYTHOS model is much clearer: The intervention that was performed on a single piece of software back then has now scaled to the entire software world. That is why seeing artificial intelligence still as a "productivity tool" is a serious mistake. This is directly an element of power.

What we call machine language has been seen as a kind of security layer for years. Code is written, compiled, translated into machine language, and it was assumed that it had become "unreadable." This assumption has quietly disappeared. Artificial intelligence can now read machine language, re-interpret it, detect vulnerabilities, and intervene when necessary. In the past, compiling was a final step. Now it is just a format change. This change, which seems small, is actually transforming the nature of software.

You write something, you compile it, it works. But that no longer means it is protected. It only means it hasn't been examined yet. While Emre was looking at the assembler, he was trying to understand it. Artificial intelligence, on the other hand, automates understanding. It doesn't get tired, it tries again and again, and it generalizes what it learns. That is why what MYTHOS does is not finding vulnerabilities one by one, but deducing the logic of the vulnerabilities. This makes the reflexes that the defense side is accustomed to meaningless.

Moreover, this is not just a theoretical discussion. It is stated that MYTHOS can find vulnerabilities even in systems that have stood out for their security for years. The fact that it can detect vulnerabilities in structures known for their robustness, such as FreeBSD, shows the seriousness of the issue. More importantly, a significant portion of these vulnerabilities are not disclosed for security reasons. In other words, unknown vulnerabilities are no longer a possibility, but an active reality.

When you add the arrival of quantum computers to this, the picture is complete. What we call encryption is slowly turning into a time problem. Data that cannot be broken today becomes solvable tomorrow. Moreover, the issue is not just the attack at that moment. Data is collected, stored, and kept. It will be solved when the right time comes.

Therefore, the issue is no longer technical. Artificial intelligence is turning into a layer that not only produces software but also analyzes it and can break it when necessary. This makes it directly an element of power. It is not the software itself, but the structure that can understand that software that has become valuable.

For countries like Turkey, there is a critical threshold here. This transformation is both a risk and an opportunity. Because the game is being reset. But those who enter this game with old reflexes will lose. It is difficult for structures that buy software from outside and build businesses on top of it to survive in this world for long. Likewise, systems that do not know what they are using and do not track which dependencies they are working with will also struggle.

What a student at Boğaziçi did by opening an assembler is being done today by artificial intelligence on a global scale. The difference is not just speed, not just scale. The difference is this: Artificial intelligence is no longer just producing. It is auditing, analyzing, and breaking when necessary. That is why it is no longer a tool, but directly an element of power.

In Brief

• The French Senate has unanimously approved a bill prepared against artificial intelligence companies "plundering" cultural works

• Claims sourced from The New York Times that Bitcoin's creator is Adam Back, and reports by Reuters that Banksy's identity is Robin Gunningham. We no longer live in an era where you can keep your identity hidden for a long time, like in the Francis Bacon era

• The memorization application developed by Milla Jovovich has caused a great stir in the software world. This means that you may no longer need to know software to make software. I experienced the same astonishment recently with software written by Murat Daltaban

• Claude can now use your computer for you

• New generation interface libraries that solve the artificial intelligence text layout problem are emerging

• Meta has developed an artificial intelligence model that predicts human brain reactions

• Claude Code is reaching a level where it can decide for itself which operations to perform

• The CEO of Salesforce has started testing the limits of package-carrying robots

• OpenAI has decided to shut down the Sora project

Things I Don't Understand

• Why traffic authorities do not apply any sanctions despite there being clear evidence on YouTube regarding the use of Tesla's FSD in Turkey

• How the fashion of holding a plush microphone has become so widespread and why it is so awkward; I block every video I see

• How Behçet Yalçın, known as "AkademikLink," can still serve as a faculty member at a university