I think it’s a pretty intuitive and natural idea to start small. My first real boss used to say: first build a Chevy, then build a Cadillac.
We are still firmly in pre-alpha era, building our Chevy, but during the first couple of rounds of discussion there were may be 50-100 tempting features that just sneaked onto the product backlog. Writing them down then was perfectly fine, as long as we didn’t also try to cram them all into alpha (or even beta).
I think the key has been finding the minimum subset of functionality that could be demoed to other people not familiar with the idea. This subset is probably half of what you think it should be, and it takes a certain objectivity and discipline to throw out all the bells and whistles. We settled on getting the basic flow of the application and a few main screens in place. The main requirement was to capture the idea , and not even try to build a prototype (that’s beta). This would be all there is in scope for alpha, and that would be locked until after we show it to a few people. So far we managed to stick to the plan pretty well.
The rule is “once we get the first round of feedback, we can unlock the scope”. And then it’s iterate, iterate, iterate.