Aydın Tiryaki and Gemini AI (2026)

When working with artificial intelligence models, the biggest misconception is assuming they know everything perfectly the first time. Sometimes, lost among massive datasets, they can confuse the realities of the modern industry with amateur league myths. More importantly, they can completely “forget” or skip some crucial topics of the subject. This article is a chronological documentary of the logical errors an AI (Gemini) fell into during a conversation that started with the question of how teams determine their football kits, and how this process transformed step-by-step into a flawless information source through the persistent questioning and eye-opening guidance of Aydın Tiryaki.

Phase 0: The “hallucination” of starting the counter incorrectly

Everything started with a technical oversight right at the first question. Instead of resetting the response counter when starting a brand new conversation, the AI began directly with the tag Response #11 due to a systemic hallucination. There were no lost messages between 1 and 10; the model simply started the counter from the wrong place while retrieving context, and this numbering error was only revealed by the user’s sharp observation at the very end of the chat.

Phase 1: The “primitive method” illusion (Response #11 and #12 – Gemini 3 Flash mode)

The process began with the user’s simple question, “How are teams’ kits selected in football?”. At this stage, the AI was operating in the Gemini 3 Flash (Fast) mode. Making a quick but superficial scan, the model summarized the process in a very generic way like “the home team chooses, the away team complies, the referee approves.” It even portrayed the coordination meeting held on the morning of the match as a primitive bargaining session where kits were chosen by fumbling through bags.

The user found this superficial explanation unsatisfactory and raised his first objection, saying, “It’s very abnormal to determine it in such a primitive way in this age, is it like this in the Super League too?” Although the model acknowledged a digital infrastructure (in Response #12), it maintained the confusion due to the superficiality of the Flash mode.

Phase 2: The biggest logical collapse: “Wearing the opponent’s shorts” (Response #13, #14 and #15 – Gemini 3 Pro mode)

Upon the user’s insistence, the model switched to the Thinking (Gemini 3 Pro) mode for a deeper analysis (Response #13). It started to recover by mentioning the FYS (Football Management System) and algorithmic color tests. However, the real disaster occurred in Response #14.

While explaining color clash crises, the model made a massive logical error and stated: “Sometimes the away team might even have to combine their own shirt with the home team’s spare shorts or socks.”

The user, as someone who masters the dynamics of the football industry, immediately intervened: “How is that possible? There are numbers, names, and logos on all shorts. Using the opponent team’s shorts would create more chaos, are you sure?” Upon this sharp objection, the model (in Response #15) admitted that it had literally hit a wall. It confessed that in today’s billion-dollar industry, this could only happen in amateur leagues, and it was only able to correct the fact that teams carry unbranded, solid-colored spare shorts and press machines with them thanks to this warning.

Phase 3: Guidance and bringing missing pieces to the table by the user

During this process, the AI did not just make mistakes; it also forgot to explain very critical topics that ensure the integrity of the subject. Those missing pieces were included in the system through the user’s visionary guidance:

• Referees’ kits (Response #17 – Pro mode): While the AI focused only on the teams, the user expanded the topic by saying, “By the way, of course, there are also the referees’ kits, they wear colors that contrast with the other teams on the pitch.” Only after this warning did the model start explaining the referees’ kit selection algorithm.
• Goalkeepers’ selection hierarchy (Response #19, #20 and #22 – Pro mode): While the model quickly glossed over the goalkeepers’ process, the user asked, “Do the goalkeepers choose their sets, or does the federation choose and notify them?” When the model replied, “The federation approves,” the user used his analytical intelligence to decipher the narrowing pool logic where the home goalkeeper and then the away goalkeeper make their selections (Response #22).
• The secret of patterned kits (Response #23 – Pro mode): It was again the user, not the AI, who discovered the fact that goalkeepers, entering the equation last as the fourth color group, have to wear complex colors to stand out from the other three groups: “That’s why there can sometimes be very different patterned things on goalkeepers’ kits,” serving a perfect assist to the AI.
• The black-and-white television rule (Response #28 – Pro mode): Just as the main article was about to be written, the user stepped in with a brilliant historical vision, saying, “What did we forget? In the past, when there were black-and-white televisions…”. Thanks to this guidance, how that old “light-dark color” rule evolved into the “grayscale” tests and color blindness standards in today’s digital systems became one of the most crucial points of the article.

Phase 4: The birth of the article and final touches (Response #29, #30 and #31 – Pro mode)

After all the rules, the IFAB hierarchy, and the historical/structural details added by the user were clarified, the user gave the instruction, “We can write the article.”

• First Draft (Response #29): The model created the first Turkish article compiling the entire process in the Thinking mode.
• The Final Human Touch (Response #30): The article wasn’t bad, but the user caught two subtle errors again: The model had used an incorrect statement like “Every football player having a different advertisement on their shorts” and gave the wrong timing for catalog registration processes. With these final warnings from the user, the model updated the text, and the article reached its flawless Turkish form.
• Closure and Approval (Response #31): With the user’s command, “Okay, now let’s write it in English,” the final English version of the article was produced. Writing the English version was the official declaration that the process had been successfully completed after all that faulty start, logical collapses, and meticulous corrections.

Conclusion

This process is the most vivid proof of how an AI, rather than producing the absolute truth on its own, can turn into an efficient “co-author” when combined with an attentive user (Aydın Tiryaki) who masters the subject, questions, adds vision, and catches logical gaps.

| aydintiryaki.org | YouTube | Aydın Tiryaki’nin Yazıları ve Videoları │Articles and Videos by Aydın Tiryaki | Bilgi Merkezi│Knowledge Hub  | ░ Bir yapay zeka hata günlüğü │ An AI error log ░ 13.02.2026

---

A Note on Methods and Tools: All observations, ideas, and solution proposals in this study are the author’s own. AI was utilized as an information source for researching and compiling relevant topics strictly based on the author’s inquiries, requests, and directions; additionally, it provided writing assistance during the drafting process. (The research-based compilation and English writing process of this text were supported by AI as a specialized assistant.)