Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
SKALE School: Confidential Transactions with Corbits
Recorded: March 17, 2026 Duration: 0:56:20Space Recording
Full Transcription
It's scale school time ladies and gentlemen welcome back for another episode
We're gonna be diving in to the new frontier of blockchain and crypto
Confidential tokens x402 and everything else in between as always with me this week
We have the man the protector the bearded green arrow mr. Sawyer how you doing man
good how are you that's an interesting interesting intro since what am I the
bearded green arrow you've always been the green arrow is this because I look
like Stephen Amell yeah it's like like one of the best compliments you could ever give me.
I know, I know, I know.
Quite honestly.
There's very few male actors that I look like,
and he's one of them.
So, thank you.
Well, after what I put you through 10 minutes ago,
you know, I've got to be nice to you.
And our special guest this week returning. This might be the first time anyone's ever returned
to scale school alexander how you doing man i'm doing great thank you for having me
uh i guess yeah it's an auspicious occasion i get to come back again
you know you know fun fact I actually think you were our first guest on Scale School, or at least our new version of Scale School.
Technically, Scale School predates our Twitter StreamYard thing, as well as technically, you know, you joining us.
But you were the first one to join us for our new version, which you set the stage because, well well following in your footsteps has been very
hard for a lot of people I think.
Well I appreciate that. That's a high compliment coming from you.
Thank you. Just compliments flying today. This is very rare for us guys. Usually we're
lying.
Yeah, you're going to leave totally fluffed up. This day is going to be the greatest
of your life.
My tires are way too big right now. We got to deflate these things. So let's get right
into things, I guess. The agenda of your economy. I know we talk about it a lot. I know there's
a lot of buzzwords. And I know there's days where we all kind of sit there and think probably
like, is this protocol really going to be it like is this really going to be the thing and i mean i go back and forth all the time i'm i'm a
very indecisive person but i think um the one thing that i am feeling very optimistic and confident
about is there are more and more people and real companies coming and building.
And I wrote this in my weekly wrap-up on, I don't know, Friday or whenever I posted last week.
I've been late a lot recently, my bad.
My wife says I'm chronically late.
Anyways, I think when new technology is being built,
it's really exciting when you've got all these,
you know,
builders that come in.
But the thing that actually moves the needle
is when the real companies come in.
Because the real companies are the teams
that get to invest capital and resources
and invest and they already have distribution, right?
It's amazing that we have all these like indie developers
and hackers that want to come in and try things.
They get things started.
But what actually moves the needle, in my opinion, is the big businesses coming in and saying, we want to be a part of this. And in 50 minutes, there's an 8,004 launch
day thing. And if you haven't seen the speaker lineup, it's stacked. It's like half Google
deepmine people. And so I just think, again, high level to kind of kick things off here.
I'm just really pumped to see the concept of agentic commerce, the agentic economy is really feeling good.
I feel like we're moving in the right direction.
Again, protocols aside, tools aside, that's all going to constantly change.
It's just the way tech goes. But I think the concept of what we're all aiming and driving toward
feels like it's truly at a, like a,
feels like we're kind of in maybe not the crossing the chasm point,
but we're, the chasm is maybe in sight.
Is that fair?
I'd say so.
I mean, I think we're in a position now where, like,
the dust is kind of settling and people are going to be making real use cases so they see not only the potential in
the longer kind of case, but also the pain points that the short term choices have made.
And it's always a trade off like you can never know the future, which is kind of just reality
of engineering. But these are kind of the interesting points where the kind of I don't
want to say the real work gets done, but some of the definitely less sexy work gets done. Fair enough, fair enough. So there's, I like that a lot. There's
kind of two key topics I want to talk about today. And they're very different, but they have a key
cross. One of them is fair meter, which fair meter is, I know you're kind of your baby,
you've driven this thing from scratch.
And I as a much worse SDK creator,
know that building SDKs that actually work is really hard.
So we'll talk a little bit about that today.
I know you guys have had a couple
of really exciting releases.
You guys have, I think we just saw today or yesterday we finally saw the uh the eip
2612 any token get integrated into coinbase so we're starting to finally see some uh new schemes
and extensions and plugins that export to which i think is really excited or exciting and obviously are exciting. And obviously you guys are one of the only, I think at this point, maybe three
full end-to-end implementations outside of the official Coinbase repo. I think the only other
one, if I'm not mistaken, that is full end-to-end like you guys is Daydreams. I think third one
has pieces, but I don't think they is full feature, if I'm not mistaken.
And then on the flip side, we're also going to be talking about confidentiality, which I believe we actually touched on briefly when you joined us last time.
We've made some strides on our side, and I've got some exciting updates to share as well.
Thoughtwork, you have like a point or something we can flip? I don't know where to start.
How about this? How about this?
How about this?
Okay.
Got a number in my head.
This is going to take way too long.
If you're within five, it's confidential.
If you're not, it's the other one.
Five?
No, it was seven, but you're within five.
So, confidential.
Start there.
That's just not how we pick numbers in the United States, but that's okay.
Normally, it's like the closest number, and it's like both people.
I digress.
Sorry.
We learn something every day.
All right.
Let's talk about confidential tokens, and then this is going to be a good segue, I think,
into Therameter because in order to make technology like this work, we need ways to consume. So
one of our long lagging maybe issues with blockchain is the lack of privacy, or maybe not
lack of, but it's this split around like,
some people love the fact that every transaction is public
and some people hate it.
And our argument is going all the way down
into the rabbit hole of privacy is a very fine line, right?
There's a lot of things that are just unknowns right now
in regards to legal compliance,
getting big businesses to use it. We just don't
know. And they're still trying to figure it out. There's governments trying to figure it out.
Right. And so while we go through all of these questionable things, one thing that we do know
for a fact is that everyday people, everyday agents, everyday enterprises all want generally the same thing,
which is they tend to want what they already have.
If you have a credit card, if you have a debit card, if you have a bank account,
if you have cash, if you have any traditional fiat payment method,
cash, right? Like if you have any traditional fiat payment method, when you go and make a payment
at a coffee shop, a bookstore, Amazon, whatever your chosen way to spend money is in your local
country currency, the person you're spending the money with can only see a small amount of
information. Usually they can see like
the last four digits of the credit card or they can see the last maybe a couple of digits of the bank account or whatever. Maybe they can see your name, they can see obviously what you ordered or
buying or consuming, whatever. A timestamp, stuff like that, the amount. They can't see what your
last pay stub was. They can't see if you went and put money into a brokerage firm
or into a centralized exchange for
crypto. They can't see all this information.
They can't see into your bank account. They can't see
your life. They can't see all the other things you purchased.
John Cork,
Mr. Student. I do not know if you saw
this, but
over the weekend,
seven major banks in the uk had a glitch you could see everybody's transactions for two like two days yeah well it was the like and i think there is a massive case
study to be seen there like i don't know i i think like
you get the actual information and stuff but yeah how like like that was like end like end
of the world type stuff you know yeah it's not ideal it's not ideal um but i mean you know to
kind of wrap this up and then we'll get some thoughts from from from around the tiles here
um you know confidentiality is just something again you know we've kind of coined this benchmark Some thoughts from around the tiles here.
Confidentiality is just something, again, we've kind of coined this benchmark the recent test,
but confidentiality is just one of these things where it is kind of a societal expectation when it comes to your finances these days.
And obviously, like, again, not everyone in the world has access to those things.
And obviously, in those situations,
maybe it's not yet a sign of expectation for them. Obviously, we hope it will be,
because I do personally believe that everyone deserves their own form of financial freedom.
My personal opinion is everyone should be able to attain the level of financial freedom they want.
If you want to debank yourself and go hardcore, by all means, go do that.
If you want to debank yourself and like go hardcore, like, by the means, go do that.
Me personally, I'm not into that so much.
And so I just want when I'm doing things on chain to be the same level as when I'm spending money in the real world.
And more so these days, I've started giving agents wallets and experimenting.
And it's kind of hard because I have to be very strategic about how I'm funding these agents
because I, you know, it's not that I have anything to hide on chain, but it's like,
you know, what if one day, yeah, I do wind up with like a really nice win on a DeFi play.
Like I don't want all these wallets linked to something that could happen in the future,
right? That's just not ideal. And so full disclosure, I really don't use DeFi too much.
I'm terrible at it. I'm a chronic loser. But you get the idea, right? This is the level
of expectation I want and what I also believe that agents need as they're moving money and
making decisions and people are putting them into business positions. If you've seen like
paperclip, you can run a full company of agents if you're giving them money and spending limits.
Do you want everyone to see
where that money is attached to? Or at least
maybe not what it's attached to, but how much they're spending?
I digress.
Falcor, Alexander,
Guy, thoughts?
I mean, I agree
completely. I don't have anything really
interesting to add on top of that besides saying that like,
I think as more and more people move on to the platforms that, you know, existing systems
have given more open data about transactions moving around, I think that's a hidden surprise
for a lot of people that aren't used to crypto and aren't used to the space.
So I think giving them an option is actually a really great thing, especially like you said, in the scope of being able to at least, you know, keep some control of your
private information relative to who you're paying and what you're paying for and that type of thing.
Fair enough. Going down that train of thought further a little bit, let's actually talk tech.
Going down that train of thought further a little bit, let's actually talk tech.
So privacy on the chain is a very nuanced topic.
There's obviously a lot of privacy protocols that have existed and privacy tokens.
And, you know, we do our best on this show to, again, we want to be able to talk about it, but obviously there's a lot of, you know, there's a lot of stuff out there that is better left to not discuss.
And so, you know, I think at a very high level, the probably two most well-known privacy tokens, at least from my knowledge, are Zcash, which is kind of going through its renaissance era these last six months, and Monero.
going through its renaissance era these last six months and Monero. I really don't know how Zcash
is doing it because I feel like it's kind of pretty much the same thing from like a high level
usage perspective and I feel like Monero got blocked everywhere. So don't really know how
Zcash is going through this but like hey good for them. I actually met, it was cool, I got to meet some of the developers
from behind the cryptographic side of their protocol
at a conference last year.
And man, those kids are smart.
I'm not even going to try to talk about what they're doing because I don't understand it.
because I don't understand it.
And that's okay.
That's okay.
That's why they're them and me.
But what I really kind of liked about
at least what's happening there is we're seeing,
we had Near on a couple of weeks ago
and they gave a pretty cool talk about
their near intense and what they're doing.
And we've seen them be one of the drivers
behind Zcash adoption.
Again, that's kind of cool, right?
Like we're seeing them and Solana
and some other blockchains kind of lean into that.
But there's still this weird,
like I think there's still this question about
does it work for an actual business?
Like if I get paid in it or I want to pay someone,
can I really use it?
And I think so far for most of the teams I'm talking to,
the answer is really not.
It just doesn't really have that viability
in your traditional business world.
And so, yeah, one of the things we've been kind of working on at scale
with our Byte protocol is what we call confidential tokens.
So we have encrypted transactions live right now
on our scale-based chain.
Conditional transactions,
we had a really amazing hackathon about,
within our hackathon a few weeks back.
Got a ton of great builders exploring different primitives
with DeFi and gaming and AI using those.
And now, we've kind of taken all these pieces together,
we've realized that we can create
what we call a confidential token.
What is a confidential token?
Well, it allows us to retain a higher level of confidentiality than you get traditionally
with a blockchain while not necessarily going all this way to full-blown, everything's private,
it has a greater chance of being compliant.
Obviously, again, something we're still testing in the field.
The way we do that is through encryption within consensus.
You can think about how Zcash does it.
Zcash does it at the protocol level.
It is a cryptographic token.
The entire chain is using zero-knowledge proofs and
cryptography to make every transfer shielded.
We do something similar, but we don't use zero knowledge proofs.
We actually use BLS threshold encryption,
which is what our core consensus at scale uses,
which allows multiple parties to come together and say,
we're going to encrypt this with a shared key.
And only if the majority of us agree,
do we have the ability to actually decrypt it.
Now where things get really cool
is our incredible engineering team has also come up with a third new byte protocol primitive,
which we call re-encryption. Re-encryption is really important because it solves this
in-between phase of what happens when we want to decrypt something and then re-encrypt it.
And historically in blockchain and maybe even outside,
you've probably really only heard of this come up
in the form of fully homomorph encryption,
which is where we do, we basically do mathematics
and changes on top of encrypted data.
The problem, it's really freaking expensive
and it's really slow and requires a lot of power.
And the only team that's really has an implementation
in production right now is Zama.
And they're only on Ethereum.
That doesn't really work for everyone else.
And so what we've been able to do
is take all these kind of core primitives,
bring them together and create a confidential token standard,
which actually allows us to, well, let's say, you know,
I've got X USDC, I can actually deposit and encrypt that USDC.
And once it's encrypted, the balance is now shielded.
It's now as I'm sending between, you know,
Falcor and Guy and all these other people,
they can move assets around as well.
And it stays fully shielded balance and amount.
All that happens by, you know,
the validators are doing that work.
And the key here is we can actually see
who's sending money to who.
We just can't see how much or what their balances are.
So again, if I'm standing in the Starbucks
and I'm ordering and Falcor's in line behind me
he knows that like i'm paying starbucks right however he doesn't necessarily see how much i'm
paying right unless he's like peeking over my shoulder which yeah um but i mean i hope that's
not i think my starbucks order could be very embarrassing.
But again, kind of coming back to this, right?
How many pumps of caramel do you get?
I drink my coffee and espresso black, sorry.
He wants to keep that private at all costs.
I'm more of a vanilla guy.
Anyways, anyways.
Right, so what's just really interesting is it's a unique way to now look at making tokens a step more toward this confidentiality, but where it actually is realistic for businesses. viewer keys right so for example let's say Falco and I are moving money back and forth and
you know I'm a business and a government entity says hey I need to see your books to see what you've done here because of xyz reason I need to be compliant what do I do they give me a key
I give them view access only to my transactions do I I want to do that? Probably not.
Do I have to do that to maintain compliance?
Most likely, and therefore,
that is a key part of a problem.
So I digress.
That's the basics of our confidential token technology.
We're going to have a lot more coming out over
the next three to four weeks on that.
Demos, documentation, information,
but wanted to give a quick intro here, a little bit of alpha for you on what's coming. Obviously, again, our, but wanted to kind of give a quick intro here, a little
bit of alpha for you on what's coming.
Obviously, again, our goal here is to talk more about agente commerce and yeah, curious,
I guess, you know, guy from your perspective, hearing this, how do you see this, you know,
potentially impacting enterprises and stable coins, impacting agente commerce?
What does this make you think about Fairmeter as well and
Corbis and how you guys are approaching
money?
Yeah, I mean, this is a really interesting
question, right? So one of the things
that we did with Fairmeter was we
kind of got this feeling really early on
that having centralized
schemes and centralized
chain interactions was going to be
just a complete bottleneck.
And so we didn't do it that way.
What we did with Perimeter is we started with a plug-in approach, kind of day zero, right?
And what this lets you do is it lets you take the stuff you're talking about, which is this really, you know, kind of strong perspective on how to do something,
but also extremely flexible and allow people to experiment with it
without requiring consensus from everyone involved in X402, if that makes sense.
Right.
And that's a huge thing because the problem is if you are trying to go ask for, let's
say, a foundation to allow you to have this new standard, then all of a sudden you're
going to be battling it out with everyone else who's got some type of skin in the game
on making sure it goes one way or the other way or whatever.
Right.
So if you wanted to do something like this with Fairmeter,
the really cool thing is you literally publish a plugin that worked as a payment handler and a facilitator handler,
and anybody who wanted to support this could just basically turn it on for any chain that they cared about
or any chain you wanted to implement this,
which is really a huge benefit for these kind of like edge case,
like improvements, these like things that aren't like maybe mainstream figured out yet, but are definitely on the edge.
It also means you can execute a lot quicker.
So it's like, you're not going to wait for permission in order to be able to say,
look, we have this new capability we've added.
We'd love people to start using it.
And then you let the market kind of go, oh, this is amazing.
Let's integrate this into everything we do, right?
So that's a big, big philosophical break from the rest of the X402 community for us.
And that's very intentional.
And I wouldn't say it's been hard because it's very easy to do, like architecturally.
But it has been like swimming upstream a bit with everyone involved.
That said, I still think to this day, after, you know, I mean,
almost a year of working on this stuff that it was the right choice. And it's great for
exactly what you're talking about, because it means that you and I could sit down and maybe over,
you know, a six pack of beers, implement this fully and get it deployed in production.
Right? Take the work you've done and make it work, which is really a benefit. I mean, it's huge. Heck yeah. I love it. I love it. That's one of my...
I think
the...
It's really interesting, and I think
this is why I think Fairmater will be
really big for this, is
when you
start to introduce
privacy
to tokens on the EVM,
something has to give.
It's really unfortunate.
Whether you're going all the way down the rabbit hole, again,
like this tech, right?
They're like hardcore privacy or you are looking at like a Zama
or you're looking even at what we're doing.
There's always one or two things that just don't exactly work.
And that's just due to how the EVM interprets information.
If you look at something like Zama, what breaks their ERC-20 compatibility
is they pass in encrypted amount fields to the transfer
because that breaks the ERCC20 transfer function signature.
For us, one of the things that we do is we do it as essentially a callback.
So our conditional transaction functionality is actually,
if you're familiar with Chainlink,
Chainlink uses a callback for most of its on-chain writes.
So you make a request to Chainlink in a callback for most of its on-chain writes.
So you make a request to Chainlink in your transaction
and then in some subsequent block thereafter,
there's a special function that the Chainlink adapter
has write access to and their Oracle system
is allowed to write to your contract,
which then executes some functionality
with the Chainlink value.
That's actually how Chainlink primarily operates.
Our conditional transaction technology works.
So you'll notice I'm saying conditional transaction,
not confidential.
That's because again, confidential tokens require
other components that actually be implemented.
It's the way that we do this is through,
again, this callback.
So how it works is I submit an encrypted transaction to scale. As part of that,
I have encrypted values that get stored on chain for one block. And I submit a conditional
transaction, which we call a CTX. And a conditional transaction to break that down into layman's terms
is basically an on-chain trigger. So in this case, it's an immediate trigger, i.e. I want to send a transfer right
away. However, I need the validators to work together and be able to decrypt this at the
right point in time to then execute. So again, we use our CTX technology, which is already built to
do this. So I submit the CTX on chain. And then what happens is at the start of the next block,
on chain and then what happens is at the start of the next block it automatically will go through
consensus and then decrypt and then execute now where it gets cool with our confidential token
specifically i know i've brought out a lot of buzzwords here is you then get to re-encrypt it
in the same block so the balances stay shielded the amount stay shielded, the amount stay shielded, nothing is leaked out. And so that is where it's gonna be,
I think Ferrameter will be really, really great
for scales confidential token
because it's not all in one block.
So if we think about an X402 transaction,
we're actually waiting for settlement
and we're waiting for verification.
And in most cases, fun fact, settlement on most
chains is super slow with X402 because block times are actually way slower than everyone says they
are. Your chain can be 250 milliseconds. I guarantee you it's going to take like one to
two seconds for that settlement to come in. At least my experience. And so for us, it actually
doesn't have a huge impact
because our block times are as well pretty fast,
but it means that settlement success
does have to technically wait for an additional block
because other like the transfers occur, right?
So again, just kind of breaking that down,
but I'm super pumped to work with you
and see how we can bring this as a set
against the perimeter because I think it's going to be,
well, the end goal is to bring this to agents, right?
We want agents to be able to move money the same way humans do.
And obviously humans as well.
Yeah.
Yeah.
No, it makes total sense, right?
Like, yeah, the settlement time thing, I think, has been understated,
like you said, so badly that I think that people are realizing they have to do like session oriented, like create, interact, wait, then pull like for results, which is kind of a funny, like we're kind of back to a lot of the like, I'm going to top up my local wallet because it's the only way I can get performance out of this thing.
because it's the only way I can get performance out of this thing.
We've done something, we've got a new payment scheme called Flex that we're working on.
And it ties into some of the stuff you're talking about, some of the stuff it's kind of orthogonal.
It would be really cool to see if there was some overlap.
But a big part of it is, yeah, when do you make a payment as an agent?
How long do you have to wait for the response?
If the agent's making micropayments, how much eventual consistency do you tolerate?
How much potential failed payments do you tolerate? Overspend do you tolerate?
There's a ton of interesting work there. And I think what it'll come down to is
on the same chain, like let's say Scale or Solana or whatever, there will be different
methodologies of payment that are completely policy-driven and related to the thing being paid for. Like I personally, if I ran a weather
server, I don't care if maybe one in 100,000 or one in 10,000 transactions overspends for what
the escrow account would have if I can give them data right away. It's no skin off my back, you
know, it's no problem, right? Whereas if you're doing some type of massive payment, like you're buying an art piece, or you're doing, you know,
something for, you know, buying automotive parts or something, then you really care, like, oh,
I need to actually make sure this goes through, right? Like, or whatever. So kind of depends on
on where on where you're willing to tolerate how what's your how quickly you need to deliver the
actual work. Yeah. Yeah. So there's a lot of room for that.
So, I think we're on kind of a really kind of fun edge
of where being flexible, having good policy control
allows for all these different use cases
to be working within the same mechanic.
And really the mechanic is what's annoying to build.
It's the headers, it's the body,
it's the responses, all this stuff.
Adding the schemes on top,
adding new payment methodologies is the fun stuff like it's once the framework's there definitely not with you there it's actually
maybe kind of a good crossover with both of our kind of core things right now which is you know
we've obviously got all of our privacy stuff uh you guys you have fair meter which is uh growing
very quickly uh we're seeing a ton of new developers pick it up, which is super exciting.
And obviously, your new Flex team.
Something just in general that I fully agree with, and I think we're at an interesting,
we're at an interesting experimental point in time where there's been a lot of talk recently
about how agents are going to make payments.
And I've been writing about this a lot. And, you know, my belief is, you know,
credit cards are never going to, um, they're ingrained in most, you know,
first world financial societies. Um, and that's fine. Right.
Like there's no issue with that saying that an agent is not ever going to be
Saying that an agent is not ever going to be given a credit card is unrealistic.
given a credit card is unrealistic.
I think the annoying part is agents need guardrails first and foremost.
And defining guardrails with stable coins is functioning a lot easier because the part that I think everyone continues to forget is that blockchains are really stinking good at this.
Like they are. And I think there's,
it's kind of funny because
blockchain has forever been this weird,
I think, technical trade-off for people
where they think they can only use it
for very certain things
and everything else has to be off-chain.
And people forget that you could just
write this logic on-chain and it works.
There's a reason people don't do this on Ethereum
because it's really expensive. But could you do it on scale? Could you do it on Solana? Could you
do it on base? And then you're like, that's not going to break, right? Your agent is not going
to accidentally overwrite logic that is by default immutable. Although maybe not so much on Solana
because they do have, I think more mutability, but at least on the EVM, it's a
little bit harder to do that.
But I digress.
The part that I'm actually more interested in here, which I love that you bring up, is
what agents are paying for?
Because this has started to become a little bit of a heated discussion where we're seeing
some bigger thought leaders in the space, some VCs, research teams, stuff like that,
kind of come in and say, hey, this agentic economy stuff is really cool,
but there's really not necessarily a ton of value in my agent using stable coins for really large payments.
And there's not value in my agents using credit cards for really small payments.
credit cards, really small payments. And when is my agent going to do one or the other?
And when is my agent going to do one or the other?
What's more valuable? And I just think it's a really interesting,
What's more valuable?
it's really interesting to me because something that V2 has, the X-402 V2 spec has that no one
has yet done from my knowledge is it can take other forms of payment that are not stable points.
It's not actually defined how to do it per se,
but it is theoretically capable of it.
Eric made sure of that in his kind of like
driving of the second iteration.
So I'm just curious if that's come up at all
as you guys are building, you know, FairMeter.
Yeah, not to jab, but it was possible in V1 too.
It's just you had to properly implement it,
which was a tough one for a lot of people.
Yeah, I mean, I know you teed me up well on that.
I think that the asset differentiation
and also just the mechanics, right?
Like sometimes you can represent, like we're currently,
you know, Solana's got a couple token standards and I'm gonna speak about Solana because I know
it best, but it's got a couple of token standards. One supported well and one's barely supported
at all. But the mechanics for sending them, receiving them are all identical, right? But
when you're talking about potential alternatives, you know, assets that maybe have slightly
different interaction methodologies. Yeah. You really need to kind of figure out,
do they fit into the existing schemes or do they,
can you extend, like you can add new ones, right?
Yeah, there's a lot of research work
that can be done there really easily
with kind of the current infrastructure, which is cool.
But I don't think a lot of it's been done yet.
Like you said, it's like, there's, you know, there's a lot of stuff with like, you know, you look at the Circle guys, what they're doing.
You look at some of the stuff people are doing with like effectively tokens over X402 where they're not even putting it.
They're putting it on chain, but it's private chain, but it allows the mechanic to work.
Like there's tons of stuff that you can do there and keeping it flexible and making it so you could potentially tilt up a facilitator that is capable of handling and of these different
schemes quite easily becomes a huge, huge boost in your ability to like adapt, right?
Which is all what I'm all about, pragmatically adapting to the problem.
Definitely.
Definitely.
I know that makes a lot of sense.
I mean, I feel like we're, it's interesting, right?
Because, you know, so much of it is driven by humans.
We're not yet at the point, I think, where we're seeing agents drive that decision-making.
There's been a lot of exploration over the last six to nine months about agents being
very capable of solving impossible math problems or science and research problems, things that
have been stuck or nothing new has been done in 50, 60, 70, 100, 200 years.
And now LLMs are actually capable, not LLMs, but themselves,
energetic loops that are capable of really high reasoning
are now solving problems and coming up with new,
literally like they're creating new forms of viable solution.
And so I think what's really exciting is I think we're kind of at that point
where there's a lot of room for experimentation and growth. And kind of back to one of the first
things I said, what I do believe is really exciting is that we have some big players involved,
right? I think it's not as relevant if you don't have Stripe involved, you don't have Visa involved,
you don't have Google involved, you don't have Coinbase involved. You don't have Visa involved. You don't have Google involved. You don't have Coinbase involved. These are companies with huge reserves of research and
personnel that actually can implement these things fairly, I mean, at this point, actually fairly
quickly. I thought Stripe was going to take a lot longer, props to them. I tested out their
machine payments last week and it was pretty good. It's not like anything
game-changing per se, but the seamlessness of how it implemented into their existing product
was good. For teams that need a high level of compliance and proper auditing and tracking for
taking payments and all that good stuff. It's a good solution. And so I think I got to the kind of piece
that led us down this track.
As agents start to pay for more things,
I think it's great that we have multiple ways to pay.
But again, I do think there's a situation where you may want to give it a credit card
with guardrails, which I still kind of feel like is more a debit card with limit split versus me,
versus stable coins. I am curious if you think we'll see any, I don't know, is it possible that
we see a net new primitive that doesn't exist yet? Do we see, is the issue not on the actual payment method?
Is the issue on real world tie-in?
Like if an agent needs to buy something, right?
We saw like 8183, ERC 8183 come out from the Ethereum Foundation and virtuals last week.
Do we need more, do we need more like middleman protocols to help just facilitate information? Is that
more relevant? Do you think in the short term? Just curious if
you have any like, this is that thing that I think I'm missing,
or you think you see is missing?
Yeah, I mean, I think I think that I mean, you and I could
talk about this for a long time. We because it's an interesting
problem, right? Like one of the things I see, and I'm not sure that this is the solution, but I definitely see the issue is, you and I could talk about this for a long time because it's an interesting problem, right? Like one of the things I see, and I'm not sure that this is the solution,
but I definitely see the issue is, you know, wallet management across a distributed organization,
doesn't matter if it's agent clusters or a big org, right now in the crypto space isn't as fluid as you'd like it to be
if you were maybe orchestrating like, let's say, a thousand agents.
Where do they, you know, like, you could build, and if anyone wants to do this, let me know,
because I don't have enough time, but you could build a really, really crazy agent-related policy-driven
faucet system. Effectively, you could have trickle-down trees or even graphs of this agent
distribution across wallets combined with basically agents requesting that things get paid for and then getting, you know, kind of confirmation. So they're not directly
interacting with the wallet, you know, data. This stuff is very clearly needed. What it'll shake out
as far as the actual implementation, I'm not sure. But I definitely see a world where agents are told
what their limits are in some type of contextual information that they get.
They're able to request payments are made
without needing to have private keys.
You know, it doesn't matter if they're smart wallets
or, you know, whatever.
They're able to ask for distribution of funds
for certain use cases with policy rails, policy controls.
This all has to exist.
But the cool thing about it is it's actually easy to reason
because if you just look at any large enterprise,
this is the exact same problem they have to deal with. You know, these guys deal with petty cash and distribution. I mean,
look at ramp, like the entire thing's built around making this automated. This isn't a new problem.
What it is, is it'll be interesting to see what the edge is between the agent and that interface
of this infrastructure. But this is a very well known, well understood, you know, you have people
with agency making actions with money
from an org. How do you do that? How do you deal with it?
Yeah, it's such a great call out and it's arguably so dumb simple. But I think it calls
out maybe the top two to three biggest points of friction, ironically, are still the same thing for the
blockchain industry, which is A, getting money in still kind of is not easy per se.
There's a lot of great companies, and I know, I will give, I will always give credit to,
I think the Solana ecosystem has done an incredible job
of this over the last six to eight months.
There's so many new teams building in this kind of like
direction of we need to make the zero to one
of getting on chain easier.
Don't love the fact that there's so many people kind of
proxying themselves as neobanks, but I like the fact that there's so many people kind of proxying themselves as neobanks,
but I like the fact that there's a lot of people trying to solve this problem of getting
money into a usable wallet is still an absolute nightmare.
And I do think there's a lot of great ramping companies, but if anyone's ever tried to put
But if anyone's ever tried to put a credit card into a ramp, you tell me how many times it actually took for that to go through.
a credit card into a ramp, you tell me how many times it actually took for that to go
through.
I'm still probably like 2% on a couple hundred tries.
It's not very good for someone living in the United States.
So I do think like that is a severe problem because, again, coming back to this idea of, well, business, right?
You're running a business.
And, you know, I kind of believe in this world where
agents will open up new jobs right like i yes are we going to probably go through this period of like
lots of people will potentially get transitioned into either no work or other roles probably in
certain industries where it's easier to let agents do things. But I do believe at the end of
the day, human CYA and human sign-off is going to be critical. And so I think it's more realistic
that that person who sits at a desk and pushes papers for eight hours a day, they might just
have a team of five or 10 agents that they're now managing and they're 10x more effective.
And this business is now 10x more effective.
Like I do see at a point that happening
because if we just fire everyone,
then like, let's be honest,
that's not going to be fun for society.
But anyways, I digress on that.
What I'm actually getting at is
if this person needs to, you know,
approve payments and do all these things
and part of that is letting the agents
have agency over them.
To your point, that average John Doe or Jane Smith working at,
the average person in the world still does not know what a crypto wallet looks like.
They don't know what an address looks like.
They don't care what a stable point is.
They don't care what an on-ramp goes.
All of that still needs to be abstracted.
All of that still needs to be abstracted.
And so I think there's that level one that we have to fix.
And then as we get deeper to your point,
how do we do wallets and guardrails in a good way?
There's a couple of teams working on this right now.
I mean, props to them.
They're in a very tough spot, I think,
because you kind of need a real organization, I think, to make it viable.
I think letting like a tech entrepreneur have 20 agents and delegate out payments is great.
But at the end of the day, like, I don't think that's your end customer. I think your end
customer is an established company who's been around for a long time. If you can get them using
agents making payments,
I think you've really broken through a threshold
that brings agentic commerce to the next level.
Just by two cents.
Yeah, I think you're right.
And I think that, I mean, look at it this way.
If Ramp tomorrow woke up and said,
hey, we're building out an agentic version
of all the workflows we've already built, right?
Let's just say hypothetically, they could make it work, right? A lot of the infrastructure they've already built. Right, let's just say hypothetically they could make it work.
Right.
A lot of the infrastructure they've spent all this time building really,
really works well with the kind of policy and guardrails and, you know,
having understanding your organizational shape is and all that stuff.
I think a big place there's going to be a problem is going to be the rate
of policy decisions and how you, what the actual quantizing is of the policy decision, how much you let out, right?
So a really good issue for us, right, going back to flex for a second because it kind of ties in, is we're at a place now where we want agents to be able to pay for their own inference with crypto.
Like basically be able to say, look, I'm coming in, I need this inference done.
I want this LLM to be able to generate this response, whatever.
Here's what I've got in my wallet, right, my allowed wallet.
I want to pay this much, this much, this much, this much, and I want it to be fast because
I'm an agent.
I need to be able to interact with these data sources, right?
Or it could be, you know, historical data on wine growing in France because I'm an agent, I need to be able to interact with these data sources, right? Or it could be, you know, historical data on wine growing in France
because I'm trying to build a real estate agent that needs to reach out, right?
The thing with X402 is we don't really have quick responses
with low millisecond turnaround on getting data back,
which means for an agentic workflow, it sucks.
Like, you're just like, okay, I'm going to wait, right?
And not to mention there are cases like with inference
where you don't know how long it's going to potentially be.
You want tokens, right?
And it maybe will cost you more,
but maybe you've got some flexibility on the policy
for how negative you go in your balance and all that.
These things are all going to get combined together
where you're going to want Web2 style,
like low latency increase in order to facilitate payment, you're not going to
be able to deal with two second turnarounds on, you know, the chain response. And also the policy
has got to be quickly decided on whether or not something could be paid through or not. So it's
this interesting convergence of like, how do we bring down latency on the actual payment? How do
we make sure that we're fast enough on the policy decisions? Maybe we've set up, you know, allowances
and escrow and that type of thing.
But there's a lot of like weird knobs that are going to get turned that I think
will get figured out as people use this more and more.
Definitely. No, I agree with you there. And I think that's a, you know,
it's funny there's a, there's a chicken and egg when you're building an agent
agent that makes payments, which is the agent itself requires an LLM. But in every case I've
that makes payments, which is the agent itself requires an LL.
ever seen, that base LLM can't yet run off X402, right? Because it has to be running constantly. I still like my thesis on like agentic agents that are more autonomous is I think in an
ideal world, every agent runs on a state of the art model and is running off of open AI
or Anthropic or XAI or Google.
But in reality, we don't have enough money to do that. It's too expensive. Paying 15 bucks
for a million token output, which is like, what, an 80-page book or something, seems like a lot.
But when you start to see these things reason and make good decisions, it adds up way too fast.
And so what we actually have is a lot of people using like open router free tier or mini max two or like cheap GLM or, you know, Kimi models, which Kimi is actually a monster.
So I mean, honestly, all these models are like pretty good, but they're not as good.
I've found when you start to get into the decision making loop around running an agent like open AI and, Anthropic, like the top frontier models have the best bank,
like pound for pound output.
And so what I think will become really interesting is
when we start to see people realize this,
people start to run the core agentic loop
on a way cheaper model,
and they start to just delegate everything out.
So the only thing the core agent loop is doing
is literally delegating tasks out,
almost like sub-agents and sub-tools constantly.
And because it's so cheap to run,
you can, if you need to call a real estate agent or you need to call
a shopping agent or a trading agent or like whatever it is, you're willing to pay for that
inference because you're saving so much on the core agent that's just running nonstop.
and you cannot pay a higher premium for that.
And you cannot pay a higher premium for that. But then to your point, it needs to be fast.
But then to your point, it needs to be fast.
You don't want your agent just sitting there waiting constantly
because it's wasting time.
Time is money.
So, yeah, I don't know.
Very, very much so agree with you, though.
The speed of this is becoming a really interesting question,
especially with these high reasoning models
because the thinking mode and the agent mode are great,
but they're also really slow,
and we like things fast.
Anyone else use chat GPT instant all the time,
even though they tell you you use thinking mode?
But yeah, sorry, to that point,
my north star for a lot of the stuff we're doing
is what would it take to build infrastructure that would allow for an agent to boot and pay for its own inference with X402 in a way that doesn't actually kill the functionality of the agent?
That's just the North Star.
It's not there yet.
So much as just being able to say, look, if I booted an agent and I needed to pay for it and I wanted to use, you know, no API keys, full pay as you go type situation, like what would it have to look like?
And that's influenced a lot of the architecture we built we're building.
It's influencing a lot of the future projects we've got kind of starting to pop up in our roadmap and some of the stuff that you guys, you know, we've been talking about already that we're about to release.
So it's a huge thing.
Love it.
No, it makes a lot of sense.
I mean, I think that is a fantastic start.
I mean, in theory, it's actually not that hard to do.
You're more so putting in a private key in almost a pre-hook.
It's like, hey, you have to use this before the agent loop
starts and then dynamically seed it.
I think the part that gets really complicated so far for me at least is um
agentic wallets which is a whole different rabbit hole um yes you can give it a private key however
giving a self manipulating agent a private key so far has been an awful idea. Yeah, what could go wrong?
What could go wrong? Sorry.
Yeah.
Everything and nothing, I guess, is kind of a good answer there.
I've been playing with a lot of agentic wallets.
I've got a lot of people asking me, hey, what's the difference
between an agentic wallet and a normal wallet?
The answer is, obviously, pretty much nothing except for the fact
that an agentic wallet is more so just capable of being called
in a way that an agent natively can write.
So not sure if you guys have had any teams playing around
that with Fairmeter.
But so far, I've tried a few decent ones.
It's definitely a growing piece to this.
But I do think that's where maybe
the limitation has started to come in
and why maybe we're seeing a resurgence
in agent cards is because I think
agentic, let's be honest,
there's like less than five good wallets
in blockchain.
Phantom set the standard.
No questions asked. Phantom has probably arguably always been the best
or one of the best wallets, I think.
I don't know, maybe you've got some other opinions,
but like there's really not that many good wallets.
Most of the existing wallets are super buggy
and super bloated.
And so I think with Argentic wallets, it's really hard
because they're all based on the same three providers.
Every Argentic wallet is pretty much privy Coinbase embedded wallets or CrossMint embedded wallets behind the scenes, which isn't a bad thing.
But they all then have the same limitations of guardrails are a little tricky.
Hooks are tricky.
You're kind of tied into a third party.
So, yeah, maybe you've got some thoughts there to tie up this trade.
Yeah, well, there's a big meta problem there too, right?
So, like, let's say that you're in a world where you are with, especially in, like, the Solana case, where you don't have access to the private key, right?
That's the reality you live in.
That's a good thing.
You can't use Xact as it currently exists inside X402. Because you can't sign the transaction to pay for it
because you don't have,
you know, you can't partially sign it
to then send it, right?
So that's going to be the new normal
is not having access to kind of secrets.
The squad's multi-sig is a really good example on Solana
of that being well done.
I think it powers some of the examples you mentioned.
I won't name names, but like the way that that works is you don't have access to critical pieces you shouldn't.
It is all programmatic.
So how does that make it work?
So that's, again, with Flex, that's a big reason we pushed the way the architecture the way we did.
Is effectively the agent in this case can generate a session key to use against an escrow account that's been funded, but it can't directly sign anything. It doesn't have any access to the escrow
itself in the case of like, you know, what are the wallets you're talking about? And I think that
that's that programmatic wallets and agent driven programmatic wallets are going to become the new
normal. I think it's just going to be how it goes. And I'm happy about it. I'm pretty excited about
it. Yeah. No, I agree. I agree fully. I think a big part, and I do think, like,
Swap is a great call-out,
selling the coolest wallet tech in the space.
They've got so much they've been building.
I think one of the really interesting things,
and I mean, we're seeing,
I think probably the best example today is Ampersand
from Edge and Node,
of a team trying to create an agentic wallet
with a human interface.
I haven't really seen anyone else pick that up
which is honestly surprising.
So I think they're thinking in the right direction
of if humans are gonna manage this,
how do humans then create this kind of trickle down effect
throughout an organization?
So I kind of expect that to be the direction
and trend that people go.
But I agree.
I think if we can make creating an agentic wallet way easier than having to go to a website and copying a command and then hoping that it finds the right open claw package and then figuring out how to pre-fund it. If we can kind of take care of all that and just make that like a five-second click a button,
start a process, and then it's done,
and then it's hooked up to your discovery
and your API abstraction and guard res and all that,
I do think we'll make a huge step forward
with downstream X402 payments, agent registration, identity.
I think it's all kind of interconnected,
so I do think that's a big missing piece right now for me.
So I like your thoughts there.
I talked to me too much today, I'm sorry.
I know we're almost at the time here.
I don't know, we're out of time.
Balbor, I will let you ask Mr.
Guy here one last question of your choice related to
anything you think is relevant,
Agenda Commerce, Barometer, Corvus.
And then I will say less
than five words to close things out.
Okay.
In your opinion, do you
think,
because I feel like as private and confidential transactions come more into place, people are going to have to choose between whether doing confidential or non-confidential transactions, right?
So you're going to have two options.
In your opinion, do you think the majority of people are going to go down confidential or non-confidential?
I think they'll do whatever is easiest and present it to them.
I think that if you create a situation, especially certain cases, where you use a business that
behooves you to provide an easy way to do confidential, then people will use it because it's easy.
And then you can help guide them towards a solution.
But I think based on the way I've seen things, I mean, you look at the US, right?
We still don't have chip and pin.
We're still chip and signature.
Seriously?
Yep.
We got it like 30 years ago.
It ain't changing anytime soon.
So that's just reality.
So yeah, I think people go with what's easy
and I think it's up to us to help them.
And also when people understand it,
when they're educated on the implications of their choices,
I think they'll make the right choice
if you give them an opportunity.
That's mind-blowing. so yeah, close us out
five words
parameter, scale
is awesome
guys, thank you so much for joining us
for another episode of Scale School
shout out Alexander thank you so much for joining us for another episode of scale school shout out Alexander. Thank you so much for joining us
You know, it's always always fun when the Corbett's guys, you know
You guys come down and we get to get to talk talk the talk. So yeah
Go go rest go rest your voice my friend
You've been I call him my little yappasaurus. He's my yappasaurus Rex
I call him my little yapasaurus.
He's my yapasaurus Rex.
He's the robot.
Thank you guys.
I am from now on.
Thank you guys so much for joining us.
And we will see you again next week.
Good boy.
