Keeping up is not the same as reading everything. There are only three kinds of news that should change what you do:
- A model got meaningfully better at something you actually do.
- A price or a limit changed on a tool you rely on.
- A tool you use changed its terms about your data.
Everything else — benchmark arguments, funding rounds, feature announcements for products you don't use — is entertainment. Enjoy it if you like it, but don't confuse it with learning.
The twenty-minute weekly habit
- Ten minutes: one roundup of what actually shipped. One source, not five.
- Five minutes: check anything affecting the tools already in your workflow.
- Five minutes: one small experiment on real work. Not a tutorial — your own task.
That last five minutes is where the learning happens. The rest is maintenance.
Watch finished workflows, not launches
Seeing someone walk through a workflow they actually use teaches more than any announcement, because you see where the effort went: the prompt they rewrote four times, the step they left manual, the thing that broke in week two. Launch posts never show that part.
Try this today
Put a recurring twenty-minute slot in your calendar and give it exactly one job: try one new thing on real work.