Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Builder Codes - Deep Dive with @RNDM_IO
Recorded: March 4, 2026 Duration: 0:28:06Space Recording
Full Transcription
γθ¦θ΄γγγγ¨γγγγγΎγγ Thank you. I'm going to go to the next video. Music I'm going to go to the next episode. I'm going to go to the next video. Music Music Music
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music Music Music Music Music Music Music Music Music Music Music Thank you. All right.
Hello, hello, hello.
Hello. Hello, hello. hello hello hello hello hello hello hello let's stop my music for a sec
perfect all right here we are thank you very much uh everyone for joining um
i believe i've tried to make a couple people co-host.
I don't know if Labs or Greg,
you've been receiving the invites.
For some reason, it isn't coming. I did.
I'm just lazy and I don't accept.
All right, perfect.
Yeah, it's just in case the main account drops off.
That way, at least the space doesn't just crash.
So yeah, we're here.
We know X can be a bit difficult sometimes.
But no, again, thank you very much all for joining.
Today's BuilderCode introduction and kind of deep dive is with Random.
And I believe behind the mic we have Vijay.
Vijay, I'll let you introduce yourself.
And then we also have another special guest from Atlas Labs.
Hello, all. happy to be here.
Yeah, so I'm Vijay, founder of Random, and we are building like agentic yield strategies and agentic worlds and everything you know built on top of different agents.
Fantastic. Very nice to have you Vijay.
And of course, here we have Greg from Atlas Labs.
Feel free to introduce yourself.
Yeah, actually, you know, it's the first time I actually got to hear Vijay's voice,
so that's super exciting.
Yeah, my name is Greg.
If you've been on a previous Aptos space, you probably know who I am.
One of the founding engineers, and I pretend to be a DevRel once upon a time
and every now and then.
So I'm here just to be the super technical guy and super excited to be a DevRel once upon a time and every now and then. So I'm here just to be the super technical guy
and super excited to be here.
Love it, love it.
Thank you very much for joining us.
Yeah, I mean, I think we've got a couple people tuned in.
I think people are kind of ready to hear you guys out.
Why don't we kind of jump right into it?
I mean, I think I'll give a quick little intro
if some of you don't already know me.
Mr. Brown, I'm the head of ecosystem here at Decibel.
And so over the course of, I guess,
the next week to 10 days,
every single day we're aiming to have a space
introducing our builders
that are essentially building on top of Decibel
or launching a product that top of Decibel or launching a product
that's supporting Decibel. So today we have Random and so they're their honorary first guest
and I hope you guys enjoy this space and we'll have open questions at the end for some of the
top questions asked by the listeners. So yeah,
let's, I guess, jump right into it, Vijay. So, you know, I guess, for the community, you know,
you describe random as like this agentic yield layer. Many protocols focus solely on creating
a better vault UI, but you've kind of prioritized automation at the strategy level.
Why is purely human-led or these kind of static management
no longer enough to capture yield efficiently
in today's high-speed perp markets?
And what really sets you guys apart?
Yep, yep. I'm glad to explain in detail about that.
So, Preface, when, you know, when we were, like,
started building this agentic strategy,
sort of like, you know, 2023,
and, you know, towards mid-2024,
this question at that time was even more, like,
you know, like, almost like a moonshot.
They were almost like, you know,
how can we trust an agent,
or how can we trust, like, you know, a bot,
and what's the difference between a bot and an agent,
and, like, why should we prefer? Humans humans are always better human traders are always better so I'm just trying
to coalesce all these questions to like a simple set of answers and like what has happened also in
2024 so if you look at it like every vault needs like a curator or the protocol itself designs its own strategy like a set of if-then-else
statements and this sort of like you know was like the benchmark so initially you had the first layer
with our way where like you know you would keep it like very very simple simple in the sense like
obviously there's a lot of complexity on interest layer interest calculations and all that but what
i meant is at the strategy level it was simple, matching the lenders and the borrowers.
Then you had Morpho with P2P.
And then as it evolved, then curators came in.
And what happened is the curators would say like,
okay, we will run our proprietary strategy
on a centralized exchange or somewhere else,
use that and make sure we deliver the best yields and so on.
But you saw what happened last
year, you know, middle of last year. So it turns out humans are not so flawless, you know. So one
part is the risk. They underestimate some of the risk. And the other part is like also the, you
know, the moral hazard, as they call it, like, you know, some of the protocols have took too much
risk or like, you know, they didn't like, you you know do that so so what we have as a base case is humans are not that perfect
and also they focus a lot on like you know the bigger fruit or the bigger parts in the value
chain so that's where our agent equals makes sense because the idea would be like with automation
we don't need like a massive team
you can do with smaller teams so obviously like you know to reach a scale you still need some kind
of bdn marketing but you can do it with smaller teams that's the first thing and this has been
proven again and again whether you look at in web 2 or web 3 agentic protocols are doing it with
smaller teams and then the next part is with Agentec E-layers,
what we are trying to do is like,
not just have pure automation as in like, you know,
if else, but also add a little bit of, you know,
like an LLM twist, like do like some critical decisions,
have like some form of learning, some form of intuition
and improve and provide yield.
So that's what we can do.
Now that's fantastic we can do.
No, that's fantastic.
I mean, thank you very much for that thorough breakdown.
And I think, yeah, I think agents nowadays are kind of almost,
you know, right now it's my full feed.
So it's how do you become more efficient?
Because if you don't use it, you're basically essentially behind.
And I think the same can be said for almost everything., comes down to even the vaults and kind of yield
opportunities
agents and
automations are the way forward
thank you very much for that, I don't know if Greg
has any input on that as well
but yeah Greg, feel free to chime
in whenever you feel like
yeah
I mean, I, so being a dev,
I was always skeptical about AI
because I remember I tried it out
like a year or two ago
and give you a bunch of hallucinations,
a bunch of random stuff that you're like,
oh no, I was doing the wrong thing.
But over the last few months, like six months or so, we've seen such great increases in
that.
You're right.
We should be admitting everything with AI.
Really, at this point, we should just be checking things.
And AI can do it way faster than us.
It can make things better than us.
It can, for like agentic trading, make decisions that aren't like emotional, right?
If you see something drop really quickly, you go, oh shit, what am I going to do?
Should I get out right now?
And maybe that's not the choice you want to make.
And then also like the agents can do microtrades and things way faster
than you could ever do it yourself.
So super excited to see what comes from it.
I'm totally with you on that.
It's funny because I'm typically one of those users that emotionally trades.
It's like the rule, never trade right before going to bed or at night. It's a no-go. It's always the worst decisions. And so exactly, you pay per hand.
And I think that's where you're right in saying having these bots that can kind of, you know,
act on emotionally, act based on data, that they can kind of ingest so much faster and so much
more than us, and creates just a more robust analysis of where
things are going to be able to execute based on on that analysis that it has it's it's it's i mean
incredible what we can do nowadays yeah totally but um yeah let's let's let's kind of hop on to
i guess um you know if you if touching more on more on the core technical unlocks that occur when an agent
manages a yield strategy on top of perps versus a user doing it themselves, for example.
Specifically, how does an agentic workflow, I guess, reduce the execution friction that
would typically prevent retail users from successfully running, you know, complex strategies.
So that's where the first part of our vault, which is Atlas.
So we had, like I said, we had Atlas in the past before we were running it on previous Perp Exchange as well.
And now we are like fully migrating it from Merkle to Decibel.
And the idea is like if you look at atena or like other
delta neutral walls what they do is like they have like a fixed uh you know a spot uh position
on one side and then i have a fixed short position on the other side and with that they try to
generate like uh it's essentially a carry trade and you're maximizing that and generating yield out of it.
Like, you know, staked yield plus, you know, funding rate.
Now, the thing what you can do with agents or what you can unlock is more complex strategies.
You know, what was reserved for, you know, structured products by, you know, bigger market makers.
So what we can do in this case is like slightly vary the leverage.
It's like slightly vary the leverage.
And on the other side, you could have like, you know,
more concentrated yield from UniV3 or equivalent protocols.
Combine that to maximize yield.
And also like try to keep, you know, like exposure,
like minimize exposure.
So this gives a little bit of a risk opportunity for the users.
So the users can potentially like, you know, suffer from del from deltas but on the other hand you could have like you
know much higher yield like greater than 25% and that's what we're doing because
there is no point in using this powerful weapon you know agents and then
delivering 5% yield you know nobody wants that so that's what we are trying
to do and that's the unlock what we what we can do with PURPS the second type of vault what we are working on is the market neutral walls so we had
it already live for the last seven months and in clay markets there was
another perp exchange and we also had it in a few other places and the idea would
be like that trades on like a lag like variations and between BTC and you know
like another asset,
like in this case, for example, APT.
And it uses that to sort of like generate like a market neutral lead.
So think of it almost like an index.
So that was our second world strategy.
But the second world strategy obviously is like a lot more risk
and more exchange different.
So these are the two offerings what we are targeting.
Amazing. Thank you very much for that. So these are the two offerings what we are targeting. Amazing.
Thank you very much for that.
I'm just curious.
I mean, you said Atlas
has already been somewhat market
proven and it's already a robust product.
Can you just tap a bit more
into that again?
Why should someone essentially come and trust it right off the bat?
So I would say, of course, users are free to try other protocols,
try other vaults and so on.
But you can look at other agentic vaults so far.
I think the top is Giza across all chains,
and it's still not cracked the 15% mark.
And it's sort of like the gold standard.
Like, you know, you've got to crack like 15, 18%
and keep the exposure really low.
So that's one thing, a higher APY.
The second thing is like, you have also like,
with this, right, you're not like locking up
for like a month or two months or something like that.
We almost have like very
minimal low uh very minimal like lockup so that's the advantage of flexibility if the users feel
like the yield is not good enough or something they can always like withdraw so we are not like
locking the users as well so those are the two advantages for like any other user to try in
and the last point of course is like then they have indirect exposure to decibel and either
either you know through the points which they have exposure or if they directly deposit on the vault
uh which will be through the decibel front and they get the decibel incentives yeah
amazing amazing no thank you very much for that that's uh quite insightful um yeah no great um
yeah i guess moving over to kind of strategy
design and kind of
I guess decibels evolution.
You know, your delta neutral
vault pairs essentially
spot longs with perp shorts to farm
these funding rates.
As the decibel ecosystem grows
and liquidity continues to deepen,
how have you designed your agents
to navigate these early
market dynamics and capture opportunities, I guess, as the order book matures? So I would say
there are two parts, right? One is the number of markets and the other would be the number of
the depth in the order book and so on. So right now we are quite limited on the decibel vaults
itself, not the Atlas vault but the Decibel vault.
So that's our target.
Our goal is to have the Decibel vault directly where users can just come in and have a full perp experience.
Where we just use the perp side and combine it with gold or other assets and so on.
gold or other assets and so on.
Yeah, as it may choose.
Like, mainly I think our target would be, like,
first would be, like, you know, the regular assets,
like, have, like, BTC, ETH and all of that,
and then, like, combine that, like, have 15, 16 assets
and use that to generate, like, a market neutral index.
Amazing, amazing.
No, thank you very much for that.
I guess kind of moving over to, you know a major pain point in defi
being you know that many delta neutral volts fail during these these volatility spikes um
and i think that's always like a big kind of concern that users may have um how how do
randoms kind of agents actively monitor and adjust leverage in real time to be able to keep this neutrality intact without, I guess,
requiring the user to manually intervene, I guess, during the market crash?
So this is the main thing, right?
That's a big advantage with agents.
So it's not just having a set of rules.
It's also doing a form of risk management
where it's handling certain crashes and so on.
But obviously, like not every edge case can be covered.
Like I don't want to come here and say like, you know,
everything is safe because as you know,
AI itself is a startup and like, you know,
Web3 is also like, you know, in general,
like Web3 ecosystem is also a startup.
So we are like building on top of a startup.
So there will always be some tail risk somewhere.
But on the other hand like
uh we are able to like mitigate a lot of it uh by simply like you know like even if there is
exposure it will not be like a complete drawdown or a complete loss of assets or like uh doesn't
require like you know extreme manual intervention so our idea would be like you know whenever there
is like some kind of a red line or some kind of a crash like you know the the it'll just simply stop of course the
edge case is like we don't know how exactly adl behavior would be with decibel and also like you
know it's quite new so those parts are still there but in general it should be like robust for most
of the cases okay now that's that that's uh good to hear i guess uh brings a bit bring some comfort to
knowing you know assets are safe and um at the end of the day that it's also able to sustain
the high volatility periods and and um stress and and rebalance i guess it's it's uh it's books um
but yeah no thank you thank you for that clarification um you know, I guess adaptability
is kind of like this real moat, right?
You've emphasized agentic workflows
over static smart contracts,
this kind of calling adaptability
being your real moat.
Since we're now live on Mainnet,
can you explain how this flexibility
has allowed random to pivot strategies
as market regimes change,
ensuring kind of your vaults remain relevant long term?
Yeah, exactly. So that's one of the things, right? Like, it's not a simple shardic strategy,
like in the sense, like it varies the leverage. But right now, I would say,
our primary focus was, you know, like, ship it, like, get it out there. We are already a little bit delayed.
So we were supposed to have Atlas Live on day one itself,
but we got a bit delayed with some of the aspects
on the dashboard side as well.
We are trying to figure out the Unreal SPNL and all that.
So I don't want to go into the technical details,
but basically that's sort of our goal,
to have it ready and have it adaptable
so that you don't need to do as much monitoring.
And that's why you use Agentic Workflows.
Awesome, awesome.
No, I appreciate that.
I appreciate that.
I guess, you know, you briefly covered around
this reward structure, you know, that Testable has
and that kind of supports your user acquisition.
And so
with you currently running
an incentives program for
your early users, can you walk us through
the current rewards
that are available to users and
how you might be able to utilize potential
AMP incentives in the future to reward
your most consistent
participants in your vaults?
So the idea would be like with Atlas vaults, there will be a smaller amount of like, you know, based on the leverage ratios,
we'll find out like the open interest and based on the open interest, whatever AMP incentives we get on the agent vault.
So a proportion of that will be aimed to distribute it downstream.
So that's one thing what we wanna do
to our atlas depositors.
In terms of the vaults through the Decibel frontend directly,
they will just be, I guess, you know,
as users who use the frontend,
they will get access directly to the AMP points.
So that part comes, I think, I guess,
directly from Decibel itself itself so those are the two
ways in which users get incentives and our goal is actually more around sustainability we want
like users to actually uh be you know happy with the product and any incentives or points are just
like a bonus to them not the other way around you know like because we don't want like users to like
you know go away as after the points or incentives to dry down.
So it's our goal would be to have like a really good product where users like our product
and then, you know, the incentives are a cherry on top.
No, I couldn't agree more with that.
That's really cool.
Let's see, I was I had a question which is off the top of my head um yeah it's kind of funny like
i whenever i go into these spaces i i don't have a finance background so i i also get lost um in
things and it's funny because we're in such a industry where we basically have become financial experts right like these are things like um
like people in finance would only talk about and we've sort of uh democratized this to the average
person which is i think really cool um so for random are are there like other strategies necessarily that we would be taking that maybe wouldn't be available to the average user that you could take now with AI that like in traditional finance you wouldn't necessarily be able to do?
Yeah, that is actually interesting.
And by the way, I don't have a financial background as well.
It's all self-learned uh i
guess like you know getting burnt in some protocols like you know depositing and then actually find
oh my god what what happened and then you know getting adl as well like i didn't know about
adl last year till i got actually adl in hyperliquid so you know learn that's the best way
to learn i guess like actually putting in capital and learning. So I would say what agents have enabled us,
like, you know, we just literally,
all I do is like, you know,
whenever like this market neural strategy, right?
You just try to look at, say,
JP Morgan or existing TradFi protocols.
I mean, sorry, TradFi companies
and then try to like find out like
what part of it can be integrated into the protocol.
And a lot of these things are already done.
A lot of financial primitives are done.
What you can do using AI
is you can sort of fit it into a specific situation here.
And I would say it's more a case of a merge.
It's not like something which is completely new
or completely unique,
but it's more like merging existing primitives into our our space and sort of like offering it
like a sophisticated way and there's still a lot of space left like for example fixed yield is one
thing uh like if you look at a traditional bank right you go there and you know like your interest
rate is going to be between two to three percent for the next i don't know two years or something
like that we don't even have those kind of simple products out here. So that's sort of
our goal. We want to make sure that we offer something which is non-existent there. Not
non-existent, but it's proved and tested in the real world and then sort of adapted to
our space. So that's why we are doing that with the market neutral strategy. And this is like actually something
which is already existing.
And like, if you look at the top,
you know, like top market makers,
they already have this form of like
a indexed strategy where like
users just deposit that.
And then like they get this kind of like
higher yield with some tail risk.
So that's sort of like,
that's how I feel it.
Like we should not like try to do something completely crazy.
And then we have that risk of strategy as well.
Yeah.
And I think it's good there where you can have like ability to set how much
risk you're willing to take.
Because obviously, you know, in the finance world,
high risk leads to high reward or high losses.
We were talking about paper hands earlier.
With perps, I also have paper hands.
I've only lost money on perps.
In fact, we had a trading competition
at Aptos Experience on Merkle,
and I lost very badly within five minutes.
Oh, yeah, I was there, actually. Oh, yeah, I remember. Oh, I lost very badly within five minutes. Oh yeah. I was there actually. Oh yeah. I remember.
Oh, I lost so badly. It was, it was embarrassing.
But it's okay, Greg, you can come and lose it now on decibel. Don't worry.
We'll support you.
It's true. I can.
I'm sure I have lots of time to lose more on decibel.
I have lots of time to lose more on decibel.
But it's nice to be able to have like that risk propensity
and be able to like simplify it in a way
where AI will do it for you.
And then like you just kind of let it go, right?
Especially because we're talking about here in perps
and then, you know, whatever else will come on decibel
and then whatever you've built on top
versus like in traditional markets, you go, whatever else will come on Decibel and then whatever you've built on top versus like
in traditional markets, you go, you find some.
