I think it's a bit early to call this as transformative as a steam engine. For starters a steam engine is, compared to vibe coding a significantly simpler concept, where you only need few parts that are available plenty.

You can explain the steam engine concept in a school and students can easily grasp this, building this yourself requires mediocre knowledge, materials and resources (money).

I think the comparison here should be the output required by the end product of the steam engine (kinetic energy) vs vibecoding. Yes most people can use the kinetic energy just like they can use the output of llm's (one of them being vibecoding), but building a steam engine is significantly easier, simpler and more scalable then producing an infrastructure that enables vibecoding.

Building a vibecoding infrastructure from scratch requires massive amounts of resources, only available to big companies. Building your own is too expensive and complex, so since there are only a handful of people (large companies) able to do this i see much less innovation on that end and much faster stagnation.

I don't disagree that it is great to quickly churn out apps (or PoC's at least) but it will mostly benefit the companies offering the infrastructure and much less humanity as a whole.