> 1) calculating square root using pure integer math

Also introducing approaches where sqrt becomes unnecessary, the fastest sqrt is the one you don't do!

> 2) search algorithms (this is a general category, path-finding is a special case)

That is a good subject indeed, mostly when dealing with pathfinding in something else than a grid.

> 3) fixed-point math (not everyone is yet aware of its usefulness), etc.

More of a micro-optimization issue, was relevant in the days of software graphics, though still occasionally useful in shaders.

On the same subject, articles on usage of floating point in situations where finite or exact precision is required would be quite relevant these days.

On AI aspects, there are two subjects I rarely see mentioned, but often (mis)used, usually by people who don't know they're using them:
- PID controllers, and the way to "fuzzy" them to make them look like natural behavior
- introduction of delays and phase tricks in algorithms to simulate natural behavior
- decision and action trees, or how to get interesting behavior in complex simulated words where direct approaches are not practical
which could make good articles I guess (the theory behind them is light, they're more about applying simple concepts from a different perspective)