Multisig - a proof of concept
First, let me introduce three rather ordinary accounts: ak_vfWbz…, ak_dQvJH…, and ak_2aDMa….
They are not very active, and not very fat, but what is interesting is the following:
30> A = <<"ak_vfWbz6Sg1iesGWtMXmgb2RXyLbsiRTjgraD8CvZoCRKtN9zPi">>.
<<"ak_vfWbz6Sg1iesGWtMXmgb2RXyLbsiRTjgraD8CvZoCRKtN9zPi">>
31> B = <<"ak_dQvJHJRRRNn1tkPincb9exxNcd4FPMBCTbFydngr4999x3LjV">>.
<<"ak_dQvJHJRRRNn1tkPincb9exxNcd4FPMBCTbFydngr4999x3LjV">>
32> {ok, ABin} = aeser_api_encoder:safe_decode(account_pubkey, A).
{ok,<<121,217,193,218,163,75,47,195,4,168,114,105,186,7,76,188,33,100,128,49,248,234,250,65,2,104,230,...>>}
33> {ok, BBin} = aeser_api_encoder:safe_decode(account_pubkey, B).
{ok,<<82,173,197,181,238,197,103,222,101,192,80,17,170,163,132,20,141,176,165,62,159,253,193,153,36,134,182,...>>}
34> AplusBBin = enacl:crypto_ed25519_add(ABin, BBin).
<<207,28,61,15,29,61,151,18,23,141,185,198,184,123,249,22,79,48,83,111,153,25,14,156,197,14,25,167,203,...>>
(aeternity_ct@localhost)35> aeser_api_encoder:encode(account_pubkey, AplusBBin).
<<"ak_2aDMabFds4dK1XLTRT26aznU42WM7AxgmYZvuLDmvw5WiA6WPM">>
It turns out the third account is the sum of the other two accounts - and this is not by accident (the odds of that happening would be astronomically small) and also if it wasn’t by construction there would not be a way to sign transactions on behalf of this account.
Taking a step back, æternity blockchain is based on ed25519 cryptography, and signing. Signatures (which is what is used to protect/verify transactions on the chain) in ed25519 is a closely related to Schnorr signatures - and this means that with a bit of elliptic curve magic it should be possible to do (n-out-of-n) multisig with normal transactions and a cleverly constructed account.
Earlier today I sat down with @dimitar.chain and we each created an account (ak_vfWbz... is mine and ak_dQvJHJ... is Dimitar’s) - only I know the private key of my account and only Dimitar knows his private key. Then, with an elaborate song and dance number that included sending 4 different partial results (two in each direction) , we managed to produce a (valid) signature of a SpendTx from our multi-sig account ak_2aDMa... - namely th_NLHUV…v1U.
It would be unfair to say that the process is simple and smooth, and I will not go into details on how to replicate this, there is a bit of work remaining before the concept is production ready for the mass market 
… But this little experiment proves that the cryptography works, and that multi-signed transactions are accepted by the network without any changes necessary.
Thanks to @dimitar.chain for helping out in the experiment. We used an extended version of the enacl library which is in a PR here - and the multisig scheme we used is outlined in this post.
