Prompts Cakmak

AI Transparency & Documentation

AI Tools & Roles

During the project, I used AI tools as development assistants. I did not use them as autonomous contributors. I reviewed their suggestions, corrected plans when necessary, tested or manually verified the results, and decided what to include in the project.

GitHub Copilot / GPT-5.2 Codex

I used GitHub Copilot with GPT-5.2 Codex mostly inside VS Code while working directly on the repository. It assisted me with codebase navigation, GitHub issue analysis, creating and switching feature branches, locating relevant frontend/backend files, debugging API integration problems, and proposing implementation changes.

I used it especially for mentorship-related frontend work, mentor capacity/profile update logic, shared goal flow investigation, follow UI behavior, social feed interactions, backend authorization debugging, seed script debugging, and documentation-related tasks.

Google Antigravity / Gemini 3 Flash

I used Google Antigravity with Gemini 3 Flash for more agentic development sessions. It helped me create implementation plans, inspect multiple files, apply frontend changes, debug authentication and persistence issues, review UI behavior, and evaluate acceptance criteria.

I used it for web location entry features, recommendation transparency on the Explore page, notification UI updates, shared goal definition flow, profile and availability persistence fixes, mentee profile routing bugs, comment like functionality, mentorship request tests, mentorship progress/timeline UI improvements, and seed persona data improvements.

ChatGPT

I used ChatGPT to organize my exported AI prompt logs into Markdown documentation and to prepare this AI transparency summary. It was used for documentation formatting and summarization, not for direct code implementation in the repository.

---

Prompt Archive

A full prompt archive is provided as a ZIP file.

The archive contains:

• raw_prompt_logs/: original exported AI prompt logs from GitHub Copilot and Google Antigravity sessions.
• curated_prompt_logs_by_issue/: cleaned and organized issue-by-issue Markdown summaries for easier review.

The raw logs are included as the primary full prompt archive. The curated files are included only to make the review process easier.

ai_transparency_package.zip

---

Human Oversight

For every AI-assisted task, I reviewed the AI-generated suggestions before using them. In several cases, I corrected the AI's initial assumptions or asked it to revise its plan.

Examples of human oversight include:

• Correcting planned HTTP methods from PATCH to PUT when the backend expected role-based PUT endpoints.
• Asking for latitude and longitude validation in location entry features.
• Requiring a clean fallback flow when location permission was denied.
• Ensuring recommendation explanations were rendered as bullet lists rather than plain text.
• Preventing invalid UI states such as displaying NaN km for distance.
• Re-testing bugs and reporting when an initial AI-generated fix did not solve the issue.
• Asking for QA-style review against explicit acceptance criteria.

The final responsibility for understanding, testing, and accepting changes remained with me.