On May 17, 2013 I finished a 2,000,000 or more digit computation of the MRB constant, using only around 10GB of RAM. It took 37 days 5 hours 6 minutes 47.1870579 seconds. You can write marvin@marvinrayburns.com for the digits.
The program I used was based on Richard E. Crandall's work:
Richard E. Crandall,Unified algorithms for polylogarithm, L-series, and zeta variants(53 pages)

As many of you know now the MRB constant = sum((-1)^n*(n^(1/n)-1),n=1..infinity).

Here are some equations involving various forms of that summation.

The first one involves convergent series and is too obvious. The others involve divergent series.

The last two, however, are new!

Let c=MRB constant and a, c~, x, and y = any number.

sum((-1)^n*(c~*n^(1/n)-c~),n=1..infinity)= c*c~.

evalf(sum((-1)^n*(n^(1/n)-a),n=1..infinity)) gives c-1/2*(1-a).

evalf(sum((-1)^n*(x*n^(1/n)+y*n),n=1..infinity)) gives (c-1/2)*x-1/4*y.

And it appears that

evalf(sum((-1)^n*(x*n^(1/n)-a),n=1..infinity)) gives (c - 1/2)*x + 1/2*a.

I want to Play with math full-time.  I posted a help me do it page on my website at marvinrayburns.com under the link titled "Help!" Any ideas on how I can gel it to go viral and maybe get some support from it?

The MRB constant can be computed in Maple by evalf(sum((-1)^n*(n^(1/n)-1),n=1..infinity)).

On my laptop restart; st := time(); evalf(sum((-1)^n*(n^(1/n)-1), n = 1 .. infinity), 500); time()-st gives a timming of 37.908 seconds.

Using the procedure posted at the bottom of this message st := time(); A037077(500); time()-st gives a much faster timing of 1.903 seconds.

My fastest timing for 500 digits of MRB comes from my...

A better approximation gives more digits of accuracy in the result per digit of precision used in the computation than a good approximation does.  I was wondering if anyone could come up with a better approximation to the MRB constant than 31/165,