I'm not seeing how this solves an NP-complete problem at all. First, the actual press release from the researcher: http://nicolau.lab.mcgill.ca/?p=1207 http://

I'm not seeing how this solves an NP-complete problem at all. First, the actual press release from the researcher: http://nicolau.lab.mcgill.ca/?p=1207 http://nicolau.lab.mcgill.ca/?p=1207 However, > We provide a proof-of-concept demonstration of such a device by solving, in a parallel fashion, the small instance {2, 5, 9} of the subset sum problem, which is a benchmark NP-complete problem. > Actin filaments exploring the {2, 5, 9} device. The movie shows a time lapse of 200 typical fluorescence micrographs of actin filaments (shown in white) moving through a network (shown in blue) encoding the SSP with the set {2, 5, 9}. Exits labeled with green numbers represent correct results, and magenta numbers represent incorrect results. They have 17 different exits. What isn't clear to me is how the exits from the device map to actual solutions or non-solutions of the subset-sum problem. {2, 5, 9} has only 8 subsets (including the empty set), so there's more exits than what I'd call "possible solutions". Roughly half of the exits map to a "correct" answer, roughly half do not. (If you ignore the "which subsets?" questions, and just as is there a solution?, then I suppose half makes sense, in that half the exits are "no solution" and the other half are "a solution exists".) But my other question, and the one I feel like neither site adequately answers, is how do you set up the device that the protein moves through? If it requires as many exits as possible solutions to the problem (i.e., one for each brute-force "attempt"), then creating the device requires as many exits — and at least as many gates, if not more; creating the device then is still hard, and has the same issues scaling that a computer does. That you can evaluate the device quickly becomes meaningless: you must build it, first. Edit: The 17 exits correspond to the sums, not the subsets. (I was also missing how they never specified what the target sum was; this is how.) So, "correct" exits are exits where there exists a subset that sums to the number of that exit. (They're solving all possible values for the sum for a given set.) As an example, exit 14 is "correct" because there exists a subset, {9, 5}, which sums to 14. Edit: Also, here's the paper[1]. This is the weirdest hyperlinking job I've seen in a while. The only link Ars seems to give to the researcher's press release is in the image credit (and even then, only to the site's home, not the release itself…). The press release only links to the paper on the page title? [1]: http://www.pnas.org/content/early/2016/02/17/1510825113.full http://www.pnas.org/content/early/2016/02/17/1510825113.full