Mill Computing's proposed architecture is more like VLIW with lots of custom "tricks" in the ISA and programming model to make it nearly as effective as the usual out-of-order execution than a "generic sea" of small CPU's. VLIW CPU's are far from 'tiny' in a general sense.