Every few weeks one of the big labs ships the thing I was about to build. How are you deciding what to build now?

Trying to ship a product while OpenAI/Anthropic/Google ship features every couple weeks that flatten whole projects.

I had an analytics feature scoped where customers build their own dashboards. Killed it, they'll just ask an AI for a dashboard in seconds.

So now I only build stuff tied to my own data and context, the bit nobody can hand me, and rent or wait for the rest.

How are you deciding what's still worth building yourself vs just waiting for it to ship?