When the difficulty is going to a level to get Bytecoin out of the hand of CPU miners ?
This is a CPU-only coin. It would never get out of CPU miners. But I can imagine people with several CPUs outdoing smaller one-CPU miners.
Thats also how litecoin was initially advertised— "cpu only". Oops. See also: https://download.wpsoftware.net/bitcoin/asic-faq.pdf
And besides, it doesn't have to have fancy hardware for it to be totally out of the hands of normal people— "cpu only" also seems to translate into botnet-herder-mostly. It's pretty easy for a "CPU mined" system to only be economically mine-able by resource thieves.
I've been studying CryptNote quite much and as far as I understand their proposition, CryptoNight algorithm is designed in a way to make the ASIC manufacturers desist from creating the proper machines, as the production should be economically pointless.
I'm sorry to cite my own post, but as you didn't reply I assume you missed this one:
I think the whole point is to make ASICs economically unviabile. It will always be possible to create a special purpose device, which is cheaper than CPU, but if the break even point is millions or billions of those devices per year, no manufacturer would even bother going into this business. I doubt you can restrict mining to CPU-only by design, but if you apply business considerations, this might actually work out.
It's not just the research, which would be costly. On the other hand, there are already billions of computers, which are suitable for CryptoNight PoW, while I'm not aware if there is any other device close to a standard CPU's speed.
Did you read about CryptoNight in general?
https://cryptonote.org/inside.php#equal-proof-of-workUnfortunately, I only understand CryptoNight on the conceptual level, so I can't validate whether the assumptions hold in 2014. However, the underlying idea is pretty simple. Just make the algo as hard to create ASIC for as possible.