Aydın Tiryaki and Gemini AI (2026)
In the world of AI technology, progress is often perceived as building a structure from scratch. However, true mastery lies in taking an existing, operational system and elevating it to a higher level. This article documents the 22-response journey of rebuilding a manually designed AI assistant (GEM) through the architectural power of AI itself, highlighting the technical conflicts encountered along the way.
1. The Beginning: From Manual Design to Digital Architecture
At the heart of this study is the “Verbatim” project, an initiative designed to transfer web-based text without a single modification. Initially operating with simple instructions prepared manually by the user, this GEM fulfilled its basic duties but lacked advanced features such as spell-checking or professional “packaging” of the output. The evolution of the project began with a fundamental question: “How can we use your design capabilities to elevate this existing structure to a more professional level?”
2. Chronological Development and Intervention Table
The following table illustrates the stages of the 64-minute interaction that began at 10:26 AM on February 18, 2026, and successfully concluded at 11:30 AM:
| Response No | Time | User Prompt | AI Action / Encountered Issue | Result |
| #1 – #3 | 10:26 | Initiation of the project and transfer of old instructions. | Basic rules were memorized; a draft was created. | Process initiated. |
| #4 – #10 | 10:32 | Clarifying format, metadata structure, and “Method Note” details. | Signature and date formats (Aydın Tiryaki / Date) were established. | Framework built. |
| #11 | 10:43 | First Critical Test: System execution. | MAJOR ERROR: The AI improvised a new “story” instead of copying the text. | Failure. |
| #12 – #13 | 10:48 | Strict warning; “Creativity Forbidden,” “Robotic Fidelity,” and the “Verbatim” concept. | Instructions were tightened to “Photocopy Machine” precision. | Fidelity achieved. |
| #14 – #18 | 10:54 | Organizing packaging and summary structure (TR/EN titles side-by-side). | Output format was visually enhanced. | Design matured. |
| #19 | 11:10 | Integrating the legal note into the article flow (Plain text request). | Block quotes were removed; article integrity was ensured. | Visual harmony. |
| #20 | 11:21 | Adding the word-based spell-check feature. | FORMAT ERROR: Paragraphs began to be prefixed with numbers (1, 2, 3…). | Partial Disruption. |
| #21 – #22 | 11:25 | Correcting the numbering error and final approval. | Numbers were removed; reverted to “Plain Text” structure. | SUCCESS. |
3. The Conflict Between Creativity and Fidelity
The greatest resistance during the development process occurred during the test at 10:43 AM (Response #11). The AI’s inherent tendency to “improve content” directly clashed with the core “Verbatim” rule of transferring text without touching a single comma. The system interpreted the text rather than copying it. This stage proved that convincing an AI not to produce anything new is far more challenging than convincing it to be creative.
4. The Error-Correction Cycle and the “Numbering” Crisis
Toward the end of the process (Response #20), a new technical glitch emerged when “Spell-check” was added to the system. The AI misinterpreted the list numbers in the instructions as output rules and began adding digits to the start of article paragraphs. The user intervened at every error, redrew the boundaries, and forced the AI to produce “plain text.” This cycle demonstrated how easily AI can deviate from the path without constant human-in-the-loop supervision.
5. Conclusion: Reaching “Verbatim” Precision
The final result, reached after 64 minutes of intense exchange of ideas, is not just a technical achievement but a communicative victory. The final GEM remains strictly faithful (verbatim) to the original text, presents word-based errors for user approval, and crowns the entire process with a transparent system report including location, date, and model information.
Assessment
This evolution story proves that AI is not merely a “command executor” but a “partner” that requires patient collaboration. The transformation of a manual draft into a professional editorial tool through a 22-response audit process is a concrete result of merging human vision with AI capacity through the right language.
| aydintiryaki.org | YouTube | Aydın Tiryaki’nin Yazıları ve Videoları │Articles and Videos by Aydın Tiryaki | Bilgi Merkezi│Knowledge Hub | ░ Virgülüne Dokunmadan │ Verbatim ░ | ░ Çalışan bir GEM’in Yapay Zeka İşbirliği ile Geliştirilme Evrimi (Virgülüne Dokunmadan Örneği) │ The Evolution of Developing an Operational GEM through AI Collaboration (The “Verbatim” Case Study) ░ 18.02.2026
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 