Changelog mornings
A new addition to my routine has been to start every morning writing out our changelog from the day before. This is mildly surprising to people who assume that an LLM is responsible for the changelog's lovely and repetitive prose. I have no deep compunction about using an LLM for such purposes, but in my experience, it is actually not particularly good at separating wheat from chaff and is more than happy to understandably assume that the 1,000 lines of code for a feature that has not yet actually launched is more important to highlight than the 20-line subcode that fixes or improves something in the editor. More than that, it's a useful centering exercise, not just to remind myself with the distance of a good night's sleep what we actually did the day before, but also what we didn't do. There is nothing quite like going commit by commit, forcing yourself to reckon with the work that really gets prioritized.
And for the most part, I'm happy with the work that gets prioritized. We spend a lot of time on paper cuts, some of which, yes, come from going too fast, but most of which come in the natural and obvious way that afflicts all applications. And we spend a lot of time on meta work too, but the private section of the changelog is dwarfed by the public one.