How are you staying in the loop on AI APIs and tooling?

It's not secret that a lot of the AI apps we're seeing built are wrappers on model apis. But as someone just exploring the space, how do you keep tabs on what's possible? By knowing what's possible and how, it usually gets my creative juices flowing in a plausible idea space. From a cursory glance, images are using dall-e or stable diffusion, chat using open AI, and now some are using replicatehq etc etc. But if there's any tips on surveying the space that would be appreciated.