In any field, you will be influenced to follow main-stream approaches. I don't see that there's any way to avoid that; you'll need to be keeping abreast of arxiv papers, updates to programming libraries, and whatever wisdom the community can accumulate. I'd say that you should embrace the community, as I've always found it much more difficult to go it alone for reasons of inspiration, motivation, and desire to feel social approval.
If you're concerned that you will miss critical insights while following someone else's approach, set appointments for yourself to check-in with how you're working. Take an hour every month or two every quarter to think through how you've been approaching your work, and how you should change.
I sometimes think about becoming part of the more mainstream AI safety community. Most recently this was triggered by thinking about taking part in 80K mentoring.
However I am partially put off by worrying about how that changes my incentives. Would I be less able to follow my own path and be funneled down the path of thinking about whatever was flavour de jour. Would I be less likely to criticise institutions approaches to AI safety if I had to worry about antagonising people and not getting funding?
So I continue to work part time but be my own master. I miss out on the community somewhat, but I think I need the freedom to go where my questioning takes me.