In this post, I've taken some endgame positions from my analysis of the recent Speed Chess Challenge match between Wesley So and Santosh Gujrathi Vidit, where you have the opportunity to find some nice ideas
Vidit was White in each of the puzzles. Enjoy!
The above positions constitute the main lessons I learned from studying the Vidit-So games in the
1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.g3 Be7 5.Bg2 0-0 6.0-0 dxc4 7.Qc2 a6 8.a4 c5!? variation of the Catalan. You can access my full 6-page analysis of the four Vidit-So blitz games from this Catalan sideline (with some interesting opening shortcuts) by
going to my Patreon site and pledging $4 or more for the month.
My analysis shows fairly well when it's appropriate to begin activating your king, a question asked by @Zeitnot17 a couple of days ago. Ultimately it comes down to whether your king will be safe or unsafe in the center, and for that, I have a rule of thumb:
If one side has more than the material equivalent of rook + knight in an attack on the king (this can include our own king), there is potential for a mating attack.
That's why, when the queens are traded but there are still many other pieces on the board (the queenside middlegame), you'll often see strong players bring up their king toward the center (e.g. Kg1-f2-e2), but not so far that their king will come under a serious attack. I could show some examples, but I think you get the idea

@ExcellentBlunderer24 asked about how to minimize blunders. Indeed, it's not enough to just look for undefended pieces (although I do recommend that to my less experienced students) - you also need to solve tactics puzzles regularly to quickly pick up on the key patterns. It doesn't matter too much which tactics program you use, as long as you solve one tactical theme at a time (e.g. 50 pinning puzzles, then 100 forks, etc).
That also covers @EOGuel's question on how to work through tactics, though in his case I'd recommend combining this with the 'Woodpecker Method', by saving the PGN of each puzzle you solve, and when you have a good number of puzzles (at least several hundred, ideally 1000+), working through the same puzzles several times to ingrain the patterns.
These two paragraphs are 90% of how I surged from 2500ish to nearly 2700 blitz on Chess.com in a matter of months, back in 2016.

Perhaps this will also help
@RoaringPawn overcome the chess plateau he talked about in his question.
I think this training method will also help @damafe to make the patterns automatic, so he doesn't have to consciously 'look for the win' and just sees the simple tactics as they come.
Now I have a chess tournament to attend, in the meantime, have a nice day!