TL;DR No, you don’t want that. You really don’t.
I understand your need for more simplicity.
But.
In the 1990’s I worked with internet forum (telnet BBS) software that was built on text files. We were continuously craving for more and better functionality. We needed to add “columns” to data. We added a TAB and then added the extra column. We needed to add them on all existing files. Then we wanted to remove (another) piece of data. We wrote an awk script to go through all the files and remove this piece of data. We needed to put the software into maintenance mode in the meanwhile because the software would see text files with a different number of fields. It was hell. We so much craved for a better system but we needed to be able to run the software with minimal resources and thus we only had a file system. We needed… a database.
However, performance is not the only issue. Text files can corrupt as well, for instance if two processes are writing to them at the same time.
There is also something called referential integrity, which guarantees that internal references exist (for instance this post belongs to topic #152787 and user #406).
And then there is transactions, snapshotting, replication, load balancing.
Sure, your database can break down. My car broke down yesterday because the overly complex ABS control software is really vulnerable to small and seemingly unrelated issues. It’s impossible to repair it by myself. I need to fork out a significant amount of cash to get it repaired. But it has so much advantages above walking everywhere. Is it unreliable? No. It still brakes and I was immediately warned by the indicator on the dashboard.
Databases were built for reliability because text files were not.