# Gems v2 Plans
Hey all, thought I'd give everyone an update on what's going on with the Gem v2 update for yfLambda. The short answer is that I'm going to be adding a new use case for existing Gems in a new NFT project I don't think anyone else has done yet.

So, what is it?

8-bit monster sprites generated by AI on the blockchain.

Now that I have your attention, let me explain.

If you look at any existing NFT project, they'll claim to have thousands or even millions of unique combinations. But if you take a close look, it's easy to see that the actual *range* of customization is quite small. You get maybe 10 choices for eyes, hands, feet, arms, etc. etc. and their uniqueness claim is simply due to combinatorial explosion.

This is evident, for example, in Cryptopunks and Axie Infinity.



They still have a similar overall shape, and it's clear that different body parts are just being swapped out.

What am I claiming to do differently?

Well, with Gems v2, which I'm referring to as Betamons for now, I've got something special, a neural network trained to output 8-bit pixel monsters. This allows me to create very freaky looking monsters that are all *completely different* from one another.

Instead of harnessing combinatorial explosion for uniqueness, I'm using the model's parameters, which have been trained on thousands of sprites to learn a very interesting function that produces very varied output.

Here are a few examples:



So that's pretty cool. But it gets better. Using a variation of this technique, I'm also able to automate *animations*.

We can get stuff like this for free:


It gets better.

One of the most exciting parts of monster collection games is the ability to merge your monsters into something new. This is easy if your NFT is modular; you can just swap in some of the parts from both parents. But doing it with a neural net is tricker.

And I've cracked that too.

So we can merge this monster:


with this monster:


to get this monster:


And the resulting monster has some traits from both, like color and shape.

Oh, and I've also figured out AI name generation.

# Roadmap
As you can see, this is very exciting. But there's still some stuff to get ironed out. Though I can essentially automate the monster generation, animation, and combination at scale at this point, there's still a few issues. The biggest is that my model has varying quality; some monsters will look more like blobs, while others will be better defined.

But, what I'd like to push for as an alpha is a dApp that will allow users to:

1. Burn their Betagems to receive these monster NFTs.
2. Allow users to buy/trade these monster NFTs directly.
3. Allow users to combine two monsters into a new one.

Ideally, I'd also like to add battling and other features down the line, but those are stretch goals.

The actual smart contract work to make this happen shouldn't be too bad. The tricky part is figuring out what information should on-chain vs off-chain.

Anyway, I'll let everyone know how things go once I get more dev work done. My plan is to whitelist the dApp first to early stakers in yfLambda, so please stay tuned.