Hazel reports · ^report(22) · 2026-10-19 · work in progress
Week in Hazel names the right author for 29 of 33 credited PRs; its errors come from names and merges
Matthew asked for a check of how Week in Hazel, Andrew Blinn's weekly digest written by AI from the Hazel repository and GitHub, credits each team member's work, with a list of corrections where it is plainly wrong. This report reads issues 001 to 006 and the three retrospective issues, as published on 2026-10-19, and checks each credit against the author of the PR or commit on GitHub.
Short answers. As of 2026-10-19: issues 001-006 and the retrospective issues -019, -060 and -100.
- Most credits are right: 29 of the 33 places where an issue gives a PR to a person by name. Issues 005 and 006 get all 17 of theirs right. The retrospective issues name no one.
- The wrong ones come from names. In issues 002 and 004 the digest calls Matthew Hammer "Matt", for six of his PRs and a commit, while Matt Keenan, also on the team, is "Matt" in issue 006. Issue 003 also gives "Matt" a PR of Alexander Bandukwala's (#2579).
- And from merges. Issues 003 and 004 credit Matthew with Alexander's colour vocabulary, because Matthew merged Alexander's configuration branch into his own branches (on 09-15 and 09-22); every one of #1591's 148 commits is Alexander's.
- The digest corrects itself, sometimes. Issue 005 corrected issue 003's credit of #2587 (Alexander's) to Matt Keenan; the errors below are not corrected yet.
- A fix for the agent: take each person from the PR's or commit's author on GitHub, never from a name in the text or from who made a merge, and use full names where two people share one.
Corrections
Each passage that credits the wrong person, by issue, with where it is, what it says (in short) and the author on GitHub.
| Issue | Where | Says | On GitHub |
|---|---|---|---|
| 002 | the story on user-defined livelits and the livelit signature | #2576 is Matt's | #2576 is Matthew Hammer's (matthewhammer). |
| 002 | the same story, on the expansion-typing review | Matt gave the closing account of #2488 | #2488 and its closing comment are Matthew Hammer's. |
| 002 | the Blackboard paragraph | Matt closed #2525 | #2525 is Matthew Hammer's, and he closed it. |
| 002 | the Fumola store-history story | Matt's September 14 commit | The Fumola commits of September 14 (605e98081c, 2b1809627b, ef85903ad9) are Matthew Hammer's. |
| 003 | the canonical-completion story, on module bodies | Matt's module-body PR (#2579) | #2579 is Alexander Bandukwala's (7h3kk1d). |
| 003 | "Name the decision, then the color" | Matthew's colour-vocabulary work merged into configuration on September 22 | The semantic colour vocabulary is Alexander Bandukwala's (commits of 08-27 to 09-01, then #1591). On 09-22 Matthew merged Alexander's configuration branch into his own branches (47cc22fab3), and said so on #1591 that day; nothing of his merged into configuration. |
| 004 | the Configuration mode story | #1591 incorporates Matthew's colour-scheme work | All 148 commits of #1591 are Alexander Bandukwala's. Matthew's part was bug reports on #1591 (09-22) and, later, his own #2613. |
| 004 | the livelit-branch story | Matt's #2597 | #2597 is Matthew Hammer's. |
| 004 | the same story | Matt opened draft #2635 | #2635 is Matthew Hammer's. |
| 004 | the Tree Care / jump-to-piece paragraph | Matt's #2602 | #2602 is Matthew Hammer's. |
How this was put together
- Each issue's online page was read as text. Every PR or issue number it mentions (160 in all) was looked up on GitHub for its author.
- A credit here is a person named as a PR's owner in the same phrase ("Name's #N", in any of its forms); there are 33 in issues 001 to 006. One of the 33 flagged by the count is a false alarm (a comment of Matthew's on Alexander's #2482, which the issue describes correctly), so 29 of 33 are right.
- The colour credits were traced through the commits: the configuration branch's history and #1591's commits (no commit of Matthew's), Matthew's merge commits of 09-15 and 09-22, and his comments on #1591.
- Credits without a PR number (a commit, a branch) were checked by hand where the text names Matthew or "Matt"; others were not.
- The data: in this page's
<script type="application/json" id="report-data">element and in022/data/credits.json; the page is made by022/drivers/gen022.py.