And maybe eventually some tiny percent of these these little things, succeeds beyond your wildest dreams, and is now critical your business, and you have to drop 10k hours slowly rebuilding it using professionals. That's what success looks like. It's also what not paying for 100 failed software projects looks like.
That's not really the point. The problem is not that you can't do something great. Every Turing-complete language has the capacity to do that. You can write a great program in Brainfuck.
The problem is that they encourage you to do something terrible. I have seen COBOL programs that I did not wish to see. Good languages make it easier to do the right thing. They provide strong type checking even for user-defined types and well-tested abstractions (like templates/generics) so that you don't have twelve copies of the same function one for each expected input, each slowly diverging and multiplying under maintenance and developing their own subtly different bugs.
The reason COBOL and Visual Basic attracted scorn and derision is that they were awful. The problem with them is quite simple: You can use them to create something improperly in 1000 hours that you could create properly using other tools in 2000 hours. That thing will then work until you have enough concurrent users to expose the race conditions, or the Access database backing it reaches the 2GB limit, or the wind blows too hard and the hamsters powering it become frightened. At which point you'll spend 10,000 hours of overtime trying to keep it from collapsing in production while the users all burn you in effigy.
And the likes of Javascript don't suck because they aren't like COBOL and Visual Basic, they suck because they are.