Thinking and Producing with Artificial Intelligence (Article 01)
From deterministic systems to probabilistic systems and the emergence of a new engineering approach
Aydın Tiryaki and ChatGPT AI (April 25, 2026)
Introduction
From the very first moment I started working with artificial intelligence, I realized something important:
I was not actually working with a piece of software.
At first glance, this may sound like a simple statement. However, for someone coming from a traditional software background, this difference represents a fundamental shift in the way of thinking. For years, we have been used to systems that operate within strict rules, are repeatable, and reliable. Artificial intelligence, on the other hand, challenges that sense of certainty from the very first interaction.
What I will discuss in this article is not merely a definition of a technology. It is a breaking point—one that you inevitably encounter when you truly begin working with AI. And without understanding this shift, it is not possible to build sustainable systems with artificial intelligence.
This text is a distilled result of my hands-on experience and extensive discussions with ChatGPT.
The Habit of the Software World
The world I come from—the traditional software world—is extremely clear and precise. There is no ambiguity. When you provide the same input, you receive the same output. This is not a preference; it is the foundation of the system.
You write a function, and that function produces the same result even if it runs millions of times. This repeatability is the greatest strength of software. Trust is built on this principle. The system does not surprise you—you know exactly what it will do.
For this reason, control is a very well-defined concept in software development. A system either works correctly or it does not. There is no gray area in between.
The First Conflict with Artificial Intelligence
When you start working with artificial intelligence, this clarity disappears. You provide the same input, yet you receive different outputs. Your first instinct is to assume that something is wrong.
You think:
“It misunderstood.”
“I didn’t explain it well enough.”
“There must be an error somewhere.”
But after repeating the same process multiple times, you begin to realize that there is no traditional error. The system is functioning—just not in the way you are used to.
This is where a critical distinction emerges: AI is not faulty; it is simply not deterministic.
Probabilistic Reality
Artificial intelligence is not a system without algorithms. On the contrary, it is built upon highly complex mathematical models and algorithmic processes. However, it is not an algorithm in the classical sense—where each step is strictly defined and guarantees the same output for the same input.
What we are dealing with here is not a rule-based mechanism, but a probabilistic system.
Given an input, AI attempts to produce the most appropriate, most meaningful, or most likely response. But this process is not based on deterministic certainty—it is based on statistical likelihood.
This is why even within the same context, small variations can lead to different outcomes. Sometimes the wording changes, sometimes the emphasis shifts, and sometimes the result takes an entirely different direction.
At first, this can feel unsettling, because it weakens the sense of control. But in reality, this is not a weakness of AI—it is its nature.
The Breaking Point
In the world of software, the fundamental rule is simple: the same input produces the same output. In artificial intelligence, this rule no longer applies. The same input produces similar, but not identical outputs.
This is not a minor detail. This difference fundamentally changes how we approach testing, quality, and even engineering itself.
If you ignore this distinction, working with AI will inevitably lead to frustration. Because your expectations will be misaligned with reality.
The Illusion of Control
One of the most common mistakes is trying to control artificial intelligence as if it were traditional software. You keep refining your inputs, rewriting prompts, trying to fix the output—but it never becomes completely stable.
Because the system you are trying to control is not inherently stable.
At this point, the approach must change. Instead of trying to control the system, you must learn how to guide it.
A New Approach: Guidance
When working with artificial intelligence, the goal is no longer to produce a single precise result. The goal is to shape the space of possible outcomes.
This is a crucial mental shift. Instead of a single correct answer, you work within a range of acceptable results.
This leads to a new kind of engineering approach. You are no longer writing software—you are designing behavior.
Prompting Is Not Enough
At the beginning, everyone focuses on writing better prompts. I did the same. I made them longer, more detailed, more explicit.
But at some point, I realized something important: the problem was not the prompt.
The problem was the absence of a system.
Single, one-off instructions are not enough to produce consistent and reliable outcomes. What is needed is structure, rules, and a repeatable framework.
This is where concepts like Gems and GPTs come into play. These are not just commands—they are systems.
A More Accurate Definition
At this point, my definition of working with artificial intelligence has changed:
It is not about asking questions.
It is about building systems.
These systems define how the AI should respond, what it should consider, and what it should ignore.
In other words, you are not requesting answers—you are defining behavior.
Conclusion
The essence of this article can be summarized in a single sentence:
Artificial intelligence is not software; it is a behavioral system.
Without understanding this distinction, any work done with AI remains random, unsustainable, and difficult to scale.
But once this difference is truly understood, everything changes. You stop producing isolated solutions and start building systems.
Final Note
This article has been prepared through the combination of Aydın Tiryaki’s practical experience and ChatGPT’s analytical contributions. The goal is to position artificial intelligence not merely as a tool, but as a new engineering paradigm.
This article is part of the series “Thinking and Producing with Artificial Intelligence.”
| aydintiryaki.org | YouTube | Aydın Tiryaki’nin Yazıları ve Videoları │Articles and Videos by Aydın Tiryaki | Bilgi Merkezi│Knowledge Hub | ░ Virgülüne Dokunmadan │ Verbatim ░ | ░ YAPAY ZEKA │ ARTIFICIAL INTELLIGENCE ░ |░Yapay Zeka ile Düşünmek ve Üretmek│ Thinking and Producing with Artificial Intelligence░ 25.04.2026
Aydın Tiryaki 