This project is designed to be consumed as a standalone data story. The presentation is the main reading source. The notebook is included as supporting evidence and reproducibility.
Open the slide deck first. It is written to be holistic and independent, meaning you can understand the story without a live presenter.
The Jupyter notebook contains the full analysis pipeline and all intermediate steps used to generate the figures and numbers shown in the presentation.
Use the notebook when you want to:
- Check exact calculations behind a chart or statistic
- See additional plots that didn’t fit into the deck
- Understand cleaning/normalization steps
- Re-run the analysis
Important methodological note: YouTube comment timestamps were not available, so time-series plots use video upload month as a proxy for when comments occurred. This is acceptable for the launch-window design but should be interpreted accordingly.
presentation.pdfthe main narrative artifact01_gpu_eda_main-report.ipynb(and/or cleaned version): analysis + figures + supporting outputs01_gpu_eda_main-report.pdf: static snapshot of the notebook for quick viewingyoutube_videos.csv: list of videos included and their metadatayt_fetch_comments.py,yt_fetch_transcripts.py: data collection scriptsyt_normalize.py: cleaning/normalization utilitiesyt_stats.py: descriptive statistics / summary helpersfigures/: exported visuals used in slides
The core message is not “which GPU is objectively best,” but how the community behaves around launches:
- Attention is highly skewed (a few videos/models dominate discussion)
- Price/value dominates conversation across generations
- Comments are consistently more positive than creator transcripts
- External context (e.g., 2021–2022 crypto boom) visibly affects discussion topics