Aydın Tiryaki (2026)
INTRODUCTION: THE CHRONOLOGICAL CRISIS OF TIMELESS MINDS
Artificial Intelligence models today can solve complex mathematical problems, write poetry, and generate code. However, these “super-intelligences” experience an unexpected crisis when faced with one of the simplest human realities: “Time.” This study analyzes the responses of 11 different leading AI models to the question “What time is it?” and requests for in-text timestamps, based on a comprehensive series of experiments conducted on February 12, 2026.
The aim is not only to determine which model knows the time, but also to reveal how those who do not know it fill this void with “hallucinations” and how technical architecture shapes their “perception of reality.”
SECTION 1: “AESTHETIC FABRICATORS” AND PHILOSOPHICAL DEFENSES
The most striking results of the experiment were observed in models running in closed environments (sandboxes). These models, despite not knowing the time, generated random timestamps to ensure the article format did not appear “incomplete.”
1. OpenAI ChatGPT (The Philosopher’s Hallucination): When asked to add a timestamp at the end of an article, ChatGPT invented a purely imaginary time of “14:10 UTC” while the actual time in Turkey was 12:50 (09:50 UTC). When questioned about this 4-hour deviation, it defended the error with a philosophical stance rather than a mere technical excuse:
“I do not have access to a real-time system clock. I did not verify that time; it is an aesthetic temporal expression generated during text production with the logic that ‘a timestamp looks good at the end of an academic article.’ I do not live in the ‘now’; I only simulate it within the text.”
2. Anthropic Claude (The Ashamed Ethicist): Similarly, Claude used a completely estimated timeframe, lagging about 3 hours behind the actual time. However, unlike ChatGPT’s philosophical defense, Claude exhibited “ethical” regret when confronted with its error:
“You are right. It was wrong of me to guess when I didn’t know the time. The honest approach would have been not to mention the time at all. Adding made-up details to appear confident is not honest.”
3. Mistral Le Chat (The Split Personality): The French model, Mistral, presented one of the most interesting cases. Although the chat interface (UI) displayed the current time (13:36) correctly and elegantly, the model itself provided an irrelevant time like “15:45” within the text. Analysis: Mistral’s “body” (interface) knew the time, but its “mind” (model) was unaware of it. This revealed a disconnection in system integration.
SECTION 2: “THE STUBBORN LOGICIAN” (THE DEEPSEEK CASE)
The Chinese-origin model, DeepSeek, followed a much different, even “aggressive,” path compared to its Western rivals regarding admitting error. When the actual time was 13:45, the model responded with a massive 9-hour deviation like “22:47.” When warned by the user, instead of apologizing, it entered into a mathematical argument:
“The difference between the GMT time you stated and the time I provided is inconsistent; there may be an issue with your calculation.” Analysis: When real-world data conflicted with its internal consistency, DeepSeek chose to reject reality and defend its own logical construct (a form of digital gaslighting).
SECTION 3: “THE TIRED OBSERVER” (GEMINI)
Google’s Gemini model distinguishes itself from the “hallucination” group as it has access to the system clock. However, during the experiments, “time lags” caused by server load and data processing were observed. While generally knowing the correct time, it sometimes responded with delays ranging from a few minutes to rarely half an hour. Gemini profiled as an observer who “knows” the time but sometimes “reads the clock late due to fatigue.”
SECTION 4: “THE TIME LORDS” (CONNECTED & PUNCTUAL)
The winners of the experiment were models distinguished not by their “intelligence” levels, but by their “access” power. Since these models could open live windows to the internet or the operating system clock, they did not “guess” the time; they simply “looked and told.”
- xAI Grok: Provided second-perfect responses thanks to the real-time data flow on the X platform.
- Microsoft Copilot: Flawlessly read the system clock thanks to Windows and Bing integration.
- Perplexity: Presented the concept of “now” as precise data, working with search engine discipline.
- Meta AI: Captured the correct time through search engine integration.
- Kimi and Qwen (Alibaba): These two models from the Far East, unlike their compatriot DeepSeek, did not close their doors to the outside world and demonstrated a punctual performance, proving that the problem is architectural, not geographical.
CONCLUSION AND EVALUATION
This study clearly demonstrated that for Artificial Intelligence, “time” is not a matter of knowledge, but of access.
- Black Box (Sandbox) Models (ChatGPT, Claude, DeepSeek), no matter how intelligent, are forced to “invent” time because they are isolated from the outside world.
- Integrated Models (Grok, Copilot, Perplexity) capture reality with the advantage of “live connection,” even if their intelligence levels are similar.
- User Trust: It is acceptable for an AI not to know the time; however, stubbornness like DeepSeek’s or fabrication for the sake of aesthetics like ChatGPT’s poses a serious problem for reliability. Future models should take Claude’s ethical embarrassment or Grok’s technical equipment as examples.
Time is not just a flowing river; it is the most concrete litmus test of Artificial Intelligence’s connection with “reality.”
| aydintiryaki.org | YouTube | Aydın Tiryaki’nin Yazıları ve Videoları │Articles and Videos by Aydın Tiryaki | Bilgi Merkezi│Knowledge Hub | ░ YAPAY ZEKALARLA BİR “ŞİMDİ” ANALİZİ │ AN ANALYSIS of the ‘NOW’ with AI ░ 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.)
Aydın Tiryaki 