Following AI without drowning

The pace of announcements in artificial intelligence is beyond what one person can read. Trying to follow everything is the surest way to retain nothing.

Three rules

1. Few sources, chosen on purpose. A handful of newsletters or blogs that explain rather than announce.
2. A fixed slot. One hour a week, not a permanent stream.
3. Try rather than read. A tool you have used for an hour is worth ten articles about it.

What stays useful

Principles outlast products: how a model is evaluated, what it costs, where it fails. Those are the questions I write down, not version names.

This list is a starting point, not a system. It will change with use.