• practical – (as @small44 mentions) build apps for yourself
• fun – build a game
• theory – do one of the coding challenges – Advent of Code, Project Euler, Codewars, Basic Computer Games, …
• starting out – Learn X in Y Minutes
• examples – Rosetta Code

Whenever I find myself wishing for a tool to do (whatever), I habitually add its description and distinguishing features to a list that I keep for this purpose.

Then, when I want to try a new language, I already have a list of project ideas.

Not every new language is a keeper for me, so the project I choose doesn't necessarily get finished in that language. That's okay, because the process still gives me the real-world experience to find what I like and dislike about the language. It also leads to improved design and faster development when I pick that project up again in some other language, because I'll have already explored the underlying issues.

I don't, i just think of app that can be useful for me

Man. It must be nice not to be drowning in 30 unfinished projects with another 30 on your mind to start all the time.

Usually, if I learn a new language, it's because it seems like the right language for whatever project I'm wanting to undertake rather than "I want to learn a language; let's see if I can come up with a project to do in that language."

I just do a random program to help me with a problem or as small task

A few ideas:

Try to create a service that lets you send a message over ActivityPub to a lemmy community.

Try to use Elixir to read and then flip the 3rd bit of an arbitrary byte stored in your system's RAM.

Try to make a simple game in c++ compiled to Webassembly to be played in a browser.

If you like Elixer you might enjoy contributing to Akkoma, which is similar to (but better than) Mastodon.

https://docs.akkoma.dev/stable/development/