It’s year-end again, and every platform has been rolling out its annual summaries. ChatGPT just released its annual report, but after reading it I felt it wasn’t very accurate -_-

Since I used quite a few large-model platforms this year, I simply did it myself—spent some time extracting the questions I’d asked on each frequently-used platform and analyzed them. The result is a personal annual report that covers every platform.

The data sources for this analysis include: ChatGPT, Kimi, Claude, Claude Code, DeepSeek, and Gemini, totaling 5,646 raw question records. After filtering and preprocessing, 5,621 valid questions yielded 8,454 analytic entries. With simple semantic analysis + clustering, the results below were obtained. The report is rendered by Excalidraw; RSS readers please click the original post to view. Heart~