Thank you. I love data. I didn’t even think of including them, it was just a fun to me exercise.
Here's a set of 3 prompts where you can pull your data. The first step is your list of projects. For the second you can pull just certain projects or say for all projects. For the third it will use the data from the second.
From there it was just a bunch of playing around with the data. :)
List every project I own. For each, return the project name, stable project ID, a brief description based only on available project information, editor URL, published URL if any, last-edited time, and available recent edit or visitor counts. Do not label a project active or archived unless that status is explicitly available. Mark unavailable fields as unavailable, and do not infer a Git repository, branch, or sync status. Provide the results as a CSV with one row per project.
For these projects: [PASTE PROJECT NAME + ID LIST], retrieve the edit history, paging backward until the source returns no older edits or you hit a documented limit. For every edit, capture project name, project ID, exact title as returned, edit type, and exact UTC creation timestamp. Preserve titles exactly, including ellipses or truncation; add a title_truncated flag when indicated. Do not treat edit timestamps as deployment times. Export labeled CSV batches, recording the source page/cursor or timestamp range and whether coverage is complete, partial, or uncertain. Compare adjacent pages and flag exact overlaps; do not silently drop records. If you cannot continue paging, state the last retrieved timestamp and the limitation rather than calling the history complete.
Using the retrieved edit-history CSV batches, create a daily edit-count CSV by project and UTC date. Deduplicate only exact same-project records matching timestamp, edit type, and exact title; retain the source batch numbers and report how many duplicate appearances were removed. Include zero-count dates only between each project's earliest and latest observed edit. Add a separate coverage summary with each project's observed date range, unique retrieved edit count, source batches, and a clear completeness status. Treat uncertain pagination as provisional, never imply missing dates mean zero edits outside the observed range, and validate the output totals against the deduplicated records.
I love the amount of data you tracked and shared in this post!
Thank you. I love data. I didn’t even think of including them, it was just a fun to me exercise.
Here's a set of 3 prompts where you can pull your data. The first step is your list of projects. For the second you can pull just certain projects or say for all projects. For the third it will use the data from the second.
From there it was just a bunch of playing around with the data. :)
List every project I own. For each, return the project name, stable project ID, a brief description based only on available project information, editor URL, published URL if any, last-edited time, and available recent edit or visitor counts. Do not label a project active or archived unless that status is explicitly available. Mark unavailable fields as unavailable, and do not infer a Git repository, branch, or sync status. Provide the results as a CSV with one row per project.
For these projects: [PASTE PROJECT NAME + ID LIST], retrieve the edit history, paging backward until the source returns no older edits or you hit a documented limit. For every edit, capture project name, project ID, exact title as returned, edit type, and exact UTC creation timestamp. Preserve titles exactly, including ellipses or truncation; add a title_truncated flag when indicated. Do not treat edit timestamps as deployment times. Export labeled CSV batches, recording the source page/cursor or timestamp range and whether coverage is complete, partial, or uncertain. Compare adjacent pages and flag exact overlaps; do not silently drop records. If you cannot continue paging, state the last retrieved timestamp and the limitation rather than calling the history complete.
Using the retrieved edit-history CSV batches, create a daily edit-count CSV by project and UTC date. Deduplicate only exact same-project records matching timestamp, edit type, and exact title; retain the source batch numbers and report how many duplicate appearances were removed. Include zero-count dates only between each project's earliest and latest observed edit. Add a separate coverage summary with each project's observed date range, unique retrieved edit count, source batches, and a clear completeness status. Treat uncertain pagination as provisional, never imply missing dates mean zero edits outside the observed range, and validate the output totals against the deduplicated records.
I will try this!
I’d love to see what you produce!