The best ML advice I ever received, from a senior engineer on my first project:

"Stop watching tutorials. Build something that breaks. Fix it. Build something harder. Repeat."

I had been stuck in tutorial hell for months — watching course after course, feeling productive but never actually producing anything.

The day I closed the tutorial and opened a blank Python file was the day I actually started learning.

Start small. Build a bad model. Make it less bad. Deploy it somewhere. Break it in production. Fix it. Build something harder.

That messy, frustrating cycle is learning. Everything else is just entertainment.