Thanks for you interest. The project scope with regards to the AI is simple (complicated simple). Every active node on the network, apart from potentially staking, hosts a local instance of an AI initialized with some randomization of predetermined variables. it then observes the chain, markets and tries to recognize patterns and eventually begin modelling and running simulations. Once it has reached an acceptable confidence level for the user (can be set anywhere between 80 and 95%) it will begin spitting out recommendations. This is the easy part...
The more complicated (and more exciting)part is that every instance of AI on the node is part of a lager AI, that being the whole network. Literally every node becomes part of the NN and participates in decision making overall. Once i solve the consensus issue, the Ai will begin controlling blockchsain events and policies. You can see the original technical paper here :- Original Technical Draft
Rather than directly integrating it from the start i kept running into issues when i first tried the concept. By definition, blockchains are sequential, AI are on the opposite side of the spectrum. It's initial development is not linear, a small change somewhere could result in massive ripples in the data set, etc, but a blockchain is 1,2,3,4,5...etc. Also, no-one has really ever published code and guides on how to quickly mine data from existing blockchains.
Eventually, i decided, why not set a road map :
1) Get data (hence the multi-wallet), it sort of solves the issue of reliance on centralized points of failure, we no longer rely on services such as explorer apis to provide information on chains which could be compromised.
2) Re-write the rudimentary data mining code in the wallet. This means that by default, when a full node running all supported chains syncs, it already creates additional database information that can be called up very simply and much faster than any api call could.
3) Begin feeding unfiltered data to the AI kernel which then processes it into acceptable input for a NN
4) We are a go, from here it's evolution and better data mining methodology.
One of the easiest things i found that made me extremely happy is how the basic code base includes unit tests, which are easily convertible for the purposes fo running simulations
In all honesty i don't know how long it will take. But with adequate funds raised, there is the potential to garner support from developers who are in it for the money. Essentially , for each aspect, i set milestones and publish them on the github repository. Developers can submit batches of code in the form of pull requests and once a goal is met each dev gets a share based on their input. For now i am a lone developer hence the idea to find ways to convince others to join up.
Another small project that the AI could get a part in is the trading interface, have an AI do your trading while you do something else (at owners own risk of course), all this is in the pipeline, but a good funding "shove" should see it come to light.
I just posted on patreon my progress on the final touches to the basic multi-wallet GUI :- https://www.patreon.com/posts/8331811
The next step is to activate the 2FA and make sure it provides secure recovery options , then maybe do some more work on getting data of the chain.
Blockchain Mechanic, thanks for the explanation, now everything looks more clear and transparent. It seems to me that the tasks set can be realized in a reasonable time and participation in Codex Crowd Fund becomes quite attractive to me and many other investors.