A second Ruby web framework is now worth taking seriously, and it exists because some people think Rails got heavy.
Rails made its name on the promise of skipping decisions and moving fast, and for a small application that promise mostly holds. Load a large number of concurrent requests, and a different picture appears. The framework does a lot of work behind the scenes to give you that convenience: loading a substantial object model, running through layers of convention-resolution logic, and all of that costs real memory and real time per request, whether or not your application needs any of it.
Merb starts from the opposite assumption. Load only the pieces a given application actually uses, keep the core small, and let convenience be something you opt into rather than something you pay for by default whether you wanted it or not.
I have not run anything of real size on it and will not pretend otherwise. What I can say is that the argument behind it is a legitimate one, and it is the same argument that shows up in every framework eventually: the thing that made you fast in month one becomes the thing slowing you down in month twelve, once you understand your application well enough to know exactly which conveniences you actually needed.
The honest tension is that Rails’ popularity is precisely the reason it can afford to be heavy. A framework with the largest community, the most plugins, and the most documentation available anywhere gets forgiven for overhead that a smaller, faster alternative would never survive being criticized for. Merb is making a correct technical argument in a market that does not obviously reward correct technical arguments over momentum.
My guess, and it is only a guess, is that the interesting outcome here is not Merb replacing Rails. It is Rails eventually absorbing whatever Merb proves works, the way an incumbent with more resources usually does to a smaller challenger with a good idea and no distribution. If that happens, Merb will have been right and will not be the one remembered for it, which is a fairly common ending in this industry.