Lei Yang

I'm personally very interested in a form of application of software.

And I think it's a big cornerstone in us reaching stage two, which means perfectly correct code.

And I think with AI, it's quite interesting because I think the thing that AI does the best is the stuff that is hard to

Produce, but easy to verify, right?

For example, proofs of software code.

Like it's pretty hard for you to prove that a piece of software code is correct.

But if you kind of draft up a proof, it's actually very easy for you to verify.

So I think with AI, with formal verification, we are on track, but still there's a long way to go.

Totally, yeah.

And I think when people really claim, when people literally mean by like roll-ups being Ethereum equivalent, it has to be stage two because in Ethereum, you don't have to trust a security console, right?

And I think that's why I think people start to realize that actually, if we treat the roll-up scaling, the roll-up century scaling roadmap as building a bunch of...

Shards that are literally Ethereum equivalent, it's actually a very hard task.

And I think five years ago, people underestimated the difficulty of this.

And I think people also kind of overestimated the difficulty of ZK proving.

So yeah, more efficient ZK proving.

So I guess a max, I think we've made good bets and bad bets.

Yeah.

1.6 giga gas then.

1.6 giga gas, yes.

I would say yes, because if we flip a switch, it's exactly the thing we're going to run in mainnet.