What worked, what failed, and what I got wrong the first three times. Published when a piece is ready rather than on a schedule, because a weekly slot is how writing gets thin.
Nothing is published yet. These are the pieces in progress, so you can see the shape of what is coming.
Crossing from the side that specifies to the side that ships. The things that turned out to be easy, the things nobody warns you about, and why the first four automations I built were all solving the wrong problem.
In the first meeting everyone likes it. In the second, someone has to give up something real. What I learned watching hundreds of companies decide whether to actually change how they work, and how it applies to shipping AI inside a team.
The filter itself, written out. What it keeps, what it throws away, and the correction loop that makes it sharper every week. You can point the same thing at your own sources.
Two years of Obsidian, most of it wasted collecting. What changed when capture stopped being free and every note had to survive a weekly pass to stay.
The tips page is already live and moves faster. Start there.