Merkle tackled the problem of brain repair and regeneration. Cryonics, he decided, rested on two assumptions: One, you could revive a frozen brain. Two, you can also revive the memory without loss. If those two problems can be solved, the rest of the body would be trivial. (You may not even need to keep your old body: molecular surgical units can build a new, younger body for you).
There seems to be a physical basis to memory. When you remember something, some restructuring happens in your brain - as if it were rewiring or reprogramming itself in the process. Memories are also fairly persistent - they last through sleep, anaesthesia, sickness, drugs, alcohol and other sorts of physical and psychological abuse - it seems reasonable to assume that it can survive deep-freezing.
The main problem with freezing and thawing an organ like the brain is that some inevitable tissue and cellular damage will occur. Merkle then worked out a scheme for using cellular repair units, based on Drexler's insight that tissue could be repaired molecule by molecule. Assuming that the total number of molecules in the human brain was 2 x 10^23, and that it would take three years to accomplish the task, then it would need a fleet of 1.8 x 10^16 of such machines to inspect and repair the brain, in effect, a total molecular overhaul. This is a large number of assemblers, more than the number of stars in our universe, but not a large quantity as far as bulk goes. Since they are so small, 10^16 assemblers would weigh only 30 grams.
As for the feasibility of his plan, Merkle followed Drexler's argument from Nature, that a proof-of-concept already existed. "When you consider that you were built by bringing each and every molecule to your brain in the circulatory system, this should not appear too infeasible."