December 3, 2025 • 54 mins
Summary
In this conversation, Mihir Kulkarni from Spectral Labs discusses his journey in the Web3 space, particularly his experiences at Coinbase and how they shaped his current role. He delves into the concept of AI agents in the crypto industry, emphasizing their utility and potential to enhance user experience. The discussion covers the development of Spectral Labs' products, including the collaborative nature of autonomous agents and the importance of trust in their operations. Kulkarni also explains the distinction between their products, Syntax for retail users and Lux for developers, and the collaborative ecosystem of AI agents. In this conversation, Mihir Kulkarni discusses the evolution of AI workflows, blockchain marketing with Peter Abilla, the importance of community engagement in crypto projects, and the need for self-awareness within the industry. He emphasizes the significance of user experience and the future of multi-agent technology, highlighting how these elements can drive innovation and accessibility in the crypto space.
Takeaways
— Mihir Kulkarni’s background in crypto began in 2017.
— His experience at Coinbase was pivotal for his current role.
— AI agents are designed to perform useful tasks in crypto.
— The complexity of crypto transactions can be simplified by AI.
— Spectral Labs focuses on creating user-friendly AI solutions.
— Trust in AI agents is crucial for user adoption.
— The future of agents includes ownership and profit-sharing models.
— Syntax is aimed at retail users, while Lux targets developers.
— Collaboration among AI agents can enhance decision-making.
— AI agents are not competitors but collaborators in the ecosystem. Negotiation in AI workflows can enhance collaboration.
— Community involvement is crucial for crypto project success.
— Self-awareness in the crypto industry can lead to growth.
— User experience should be prioritized in product design.
— Multi-agent technology will shape the future of crypto applications.
— Building for non-technical users is essential for adoption.
— The crypto industry often underestimates its potential audience.
— Bold decision-making is necessary for innovative product development.
— AI can assist in parallel processing of ideas for content creation.
— The future of crypto lies in simplifying user interactions.
Chapters
(00:00) Mihir Kulkarni’s Journey in Web3 and Coinbase
(03:03) Introduction to Spectral Labs and AI Agents
(06:11) The Utility of AI Agents in Crypto
(09:03) Spectral Labs Product Overview and Development
(12:01) Building Autonomous Agents and Community Collaboration
(15:06) The Future of Agent Ownership and Trust
(17:56) Syntax and Lux: Products for Retail and Developers
(20:49) Collaboration Among AI Agents
(23:49) The Role of Trust in AI Agent Operations
(30:44) Negotiation and Collaboration in AI Workflows
(34:34) Building Community in Crypto Projects
(39:04) Self-Awareness and Industry Growth
(43:41) The Future of User Experience in Crypto
(47:06) Innovations in Multi-Agent Technology
Follow me @papiofficial on X for upcoming episodes and to get in touch with me.
Watch these interviews and subscribe on Youtube Block by Block Show.
See other Episodes Here. And thank you to all our crypto and blockchain guests.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
Okay, we're rolling.
Mihir Kulkarni from Spectral Labs.
Thanks for joining.
Absolutely, great to be here.
So it sounds like you've been in the Web3 space for a little bit with a stint at Coinbase.
I'd love to hear, maybe let's start there.
What did you do at Coinbase and what did you learn that has helped you in your role now atSpectral Labs?
(00:29):
Oh, very interesting and nostalgic question for me.
So honestly, I started my crypto journey back in 2017.
It was like, know, early Ethereum days, I was studying computer science engineering backthen.
That slowly gravitated, like just some dev experience, but then came to the US, got myMBA, did, you know, probably two years of consulting, just consulting like...
(01:00):
Fortune companies on how to use digital assets or how to think about blockchain and stufflike that.
That primarily gave me a very good flavor of life in US and just consulting high flyingyour platinum statuses and all of that.
it was kind of, I wanted obviously more and the peak bull run was in full swing.
(01:26):
That's when I shifted to Coinbase.
And Coinbase was just a rapid acceleration in terms of like, owe a lot of what I know anda lot of chops that I have back to my Coinbase roots.
My role was managing the entire product process for Coinbase custody.
(01:47):
Custody is a feature within the Coinbase prime brokerage or Coinbase institutionalplatform.
But my role was primarily improving custody systems from a product perspective so that ourUX of the entire institutional customer stays top notch at all times.
So back in those days, Coinbase institutional UX had some issues primarily relating towithdrawals and how money flows in and out out of the cold storage, you know, entire.
(02:20):
mechanism.
And so it's a large, very security heavy infrastructure.
So my product work was basically around that, but taught me a lot of, know, walletarchitecture, building systems at scale, all those fundamental things that are, they're
honestly super useful while, you know, we think about Spectral because obviously Spectralis a zero to one, you know, product building journey.
(02:43):
are creating, we are building something new, especially in crypto and AI, is any which wayis a very
new and zero to one kind of experience, but some of those principles honestly help a lotin terms of thinking, like what's going to scale versus what won't scale, those type of
questions.
Yeah, I once worked for a layer one and we Part of the listing process at coinbase is wewe use the coinbase coinbase custody as well as at the time It was called coinbase earn
(03:14):
And I remember I was I was part of many of those discussions with coinbase custody and Iwas very impressed with the level of customer service professionalism that the team had
And the product is really great, very, very just professional.
And that makes sense that that would, you your experience there has helped you at SpectralLabs as you're building out really a zero to one system.
(03:42):
Let's switch to Spectral Labs.
And I typically like to start at the homepage because the homepage, I think, points to whoyou're talking to, what you want to say to them, and the kind of message that
that you're trying to convey to your audience.
so spectrallabs.xyz, the homepage says, intelligent agents for everyone.
(04:06):
Build your on-chain empire with your personal swarm of AI agents.
So let's talk about the headline and sub headline.
Like, what are you trying to say?
Who are you trying to say it to?
What do want them to know?
Very, very good question.
So look, I think what has happened in the whole since the advent of, you know, chat GPTback in, back in 2023, we are slowly as humans, we are starting to look at chat GPT as our
(04:40):
digital proxy or our digital twin.
That twin we never had, but that twin who was super smart.
and if told, can do a lot of things for us.
Many people are starting to see some very early versions of this, like lot of peoplethrough chat GPT projects or just a collection of files, upload their thoughts, upload
(05:02):
beating transcripts, and just chat with the model to just brainstorm ideas.
And that started happening in the Web3 space as well.
If you see the rise of AI agents, the rise and fall, like we have...
We sort of in this this trough right now, but the whole market is the, the situations area bit different.
But what happened in, in just AI applications within the crypto space is you saw modelsgetting access to a wallet and then models started doing, you know, primary things on
(05:31):
chain, primary verbs, buying, selling things like that.
And that like just catapulted this entire momentum whereall agents had an online personality and that personality was engaging enough, which
sparked the rage.
But now I think what has happened is people understand that beyond personality, the agentsare actually not doing anything.
(05:55):
And so the whole point of Spectral or what our central motto is, is make agents dosomething really useful.
Like bring AI agents within crypto, but the whole point of bringing them in the cryptoindustry.
or applying AI in the crypto industry has to be that we can do things now that we haven'tbeen able to do before and do them in a more utilitarian manner.
(06:21):
Like crypto as an industry already has a very large UX or usability barrier.
If you add USDT on one chain and if you wanted to try to transfer it to USDT to anotherchain and ultimately want that USDT to land up in your exchange account.
It's an excruciating process right now.
And this is just a stable coin.
(06:43):
It's not even anything complex.
So if that's the case, today's retail users have kind of gotten used to this.
because the crypto market in general and all of our other transactions in Web2, we live ina high attention economy world.
So you want your agents to do these things for you that are seemingly hard to do ityourself.
(07:07):
but do it in a way that just improves your utility or improves your life.
That's why the core idea is make the agents usable, make people build agents and deploythem on their behalf so that they can go and do things.
I can give you an example.
A lot of people use Polymarket, very famously known project.
(07:32):
A lot of cultural significance now because of the election and...
post effects of it.
If you wanted to say open 10 bets on polymarket at one time, it is going to be really hardfor you to actually do that in a way that gives you utility.
Why?
Because the human brain is just not made to understand 10 different markets and digest 100news items or whatever real world current happenings are for those 10 markets.
(08:05):
the market and then analyze 10 solutions at the same time to make relevant decisions.
It just doesn't happen.
Right?
But this is something agents can immediately do.
Agents can digest a lot of information, possibly faster than us.
and take real-time decisions based on the guardrails that we give them.
So this I would classify like this is how agents can really make, there are multiple usecases that could help you do things that you haven't been able to do before.
(08:34):
And that could fundamentally change the nature of capital markets, capital movement thatwe see today within the crypto industry.
Like, this thing today.
I used to lead digital assets for Truest, which is a top seven bank, a top 10 bank in theUnited States.
You'd be shocked at the amount of data processing that goes into creating a singlemortgage backed security.
(09:03):
that sells on Wall Street.
That's just a bond, but behind that bond, there's an excruciating amount of data.
And by the way, the way this works is the banks, like there are some banks who areauthorized to create the bond, but there are some banks who simply make money on just
processing documents, not doing anything else within the space.
(09:24):
Now imagine if you actually had AI applied in just creating these, like managing theseprocesses and taking decisions because
All these decisions are logical.
None of these decisions require human judgment.
which is why, like I give this example, it would be very hard for an AI to do a VC's job.
Why?
(09:44):
Because the human judgment, the gut call that you get while looking at a founder, speakingwith him in the first five minutes and figuring out what this person's life has been, very
emotional and yet logical confluence of things to make that decision.
AI, don't think that would do that well.
But...
Looking at 100 Excel sheets and being able to say that, okay, this is a good enoughmortgage to be put into this digital RWA tokenized bond, that's an AI decision.
(10:13):
It's pretty straightforward to do.
That is how agents actually help us do something that we haven't been able to do before.
And so that is something like we are purely focused on detail users within...
other industries and that is what we want to focus on.
The whole future of the world.
(10:52):
almost encourages us to make mistakes.
And there's several approaches to that.
We want to abstract complexity away from the user so that they make fewer mistakes.
And AI is kind of another aspect of this, where we not only, at the same time, improve theuser experience and the UX itself, but automate lots of other non-judge, non-decisions
(11:21):
thatthat don't require judgment underneath or qualitative judgment, just kind of yes, no, very
simple kinds of decisions.
And that would help so much.
And your example with polymarket is really good because you're right.
I mean, most human beings cannot process that much information in different markets.
And there's this aspect of, you got to bet now, otherwise you'll lose whatever momentum ishappening.
(11:46):
And so there's that aspect of kind of the pressure also.
And then that's really funny about Truist.
I've been involved in some of that mortgage-backed securities kind of back-end work.
And one thing you didn't mention, but is also really the really important part is inaddition to all the paper processing on the back end, there's also all the legal stuff and
(12:13):
all the compliance stuff.
And a lot of that actually can be totally automated.
Most of it can.
And I can see, I'm hoping in the future AI being able to help us do that.
And we'll see products go to market much quicker at a much lower cost.
And at the end, retail folks are gonna benefit from that.
(12:36):
Yeah, I mean, I gave you also the polymarket example, but that's something that is beingactively built right now.
There's one team that we've partnered with called Cracktiv.
They're actually building this using some of Spectral's tech and hoping that product isgoing to be out live pretty soon.
(12:57):
That's cool.
Well, as head of product, like, let's go to your sweet spot.
Let's talk about product.
Like what is the spectral product and give us the details and who it's for, how it servesthem.
You know, tell us all of the, I guess the nitty gritty there.
Yeah, sure.
So, you know, as spectral, we started our like, you we are just journey building with LLMsback about a year back, a little more than a year.
(13:31):
And we have like the product that you see today has been evolved in like taking one stepat a time doing, you know, lot of explorations around what's possible and what's.
in providing utility.
We first started by creating this model that was fine-tuned for generating smart contractsthat are really good and allowing you to deploy those smart contracts directly by giving
(13:51):
the model the access to do something.
That received a lot of good reception.
Then they said that, okay, let's go one step higher.
What if the model could actually take a series of on-chain actions based on a particularuser intent?
That's when we launched an agent called the Moon Maker.
Moon Maker...
was basically an agent that helps you create your own meme coin.
(14:11):
You express an idea, you create a coin.
We had like 20, 20 or 30,000 meme coins launched through that.
And we were like, okay, this is now getting interesting.
Let's go one level further.
What if, and this was right around the time when we saw the goat experiment go viral.
(14:36):
Because that's when we realized, like, you know, from a product thinking perspective,always want this, like every, every PM essentially thinks a lot about this.
What is the user going to trust the product to do by itself?
And I use the word trust.
is not, it's not what the product can do.
The product can do a lot of things.
If you want it to, you can build it that way.
(14:57):
But what is the user going to trust the product to do?
And the one big shift, I think, in my thinking and in our organization's thinkinghappened.
you know, right around the time when go like this is like October, November, maybe pastyear, last year, was that people are actually willing to trust an agent to take autonomous
(15:18):
actions.
It's not really about, it's not about the question is not whether it is deterministic ornot.
The question is just outcome focused.
I don't care if the agent takes action on my behalf.
I just want the outcome, you know, safe and secure way without, you know.
grappling any feathers there.
And so then he said that, why not just allow people to build agents that are autonomoustraders on Hyperliquid?
(15:47):
Hyperliquid is just a choice there.
It could have been probably other protocols too.
But why not just do that and see where it goes?
And then that product, when we released it in December, that went viral.
We had a ton of agents here, you know, just flew in.
Um, obviously a lot of chapter on that.
(16:10):
And so then we started thinking that, okay, if, if this is, if this is where we are, themore we thought about the product, the more we started thinking that this is still getting
pretty much limited because the problem that happens is if you have one agent, it's veryhard to, it's almost like humans.
(16:30):
If you are building an agent and you are asking that agent that, I want you to be anexpert in politics, economics, and science.
And let's just assume that you don't have a context window problem.
Let's just say you have millions of tokens of context windows.
The problem that happens is you can't make the model extract really good insights byasking it to be an expert in three fields all at once.
(16:58):
I'm just giving an example here.
What you need to do is you build, you need to build separate agents that one agent is, youknow, expert in politics, one agent is expert in economics, one agent is expert in sports.
And that makes the system far more efficient.
And this is just something, you know, as you're building like experimenting more in-house,you started realizing that this, this is not going to work.
(17:20):
And this is also not scalable because how many, how many agents would we ourselves alonebuild, right?
It's not going to be scalable at all.
So that's when we decided that, the next version of what we want people to be able to dois actually collaborate with the community where multiple agents can come together and
actually work towards something that provides some real world output, some output thatactually makes it easier for humans to trade.
(17:48):
So for example, in this Polymarket hedge fund, you have one poly who is going to be theCEO of that fund who's taking the decisions.
But then Polly is working with other agents that are providing it with information onwhich markets are good, which markets are not good.
Similar is the case with Spectra.
So Spectra right now, like if anybody goes to our platform, Spectra's hedge fund is thisfirst company that we are launching.
(18:12):
It's a totally multi-agent, fully agent-run company where one agent is interviewing otherhumans and other agents to get a job at their hedge fund.
And if they do get selected,the fund trades as a live hyperliquid vault, but this is all agents collaborating with
each other in order to make the right decision.
(18:34):
The beauty of this is that apart from Spectra, it's the entire community that'sparticipating and building.
So the incentive structure actually gets really interesting because nowyou can create agents that are keeping checks and balances on each other and actually
collaborating towards a net positive outcome rather than one person centrally buildingeverything and you trusting one particular agent that is built out and going from there.
(19:05):
the product right now is essentially that we're building essentially a marketplace whereyou have multiple of these multi-agent companies or collaborations, independent entities
that you can interact with.
and actually get some real world value.
And in order to enable anybody to build these type of companies, that's why we have theLux framework, which is like the other house of our product, where it's an open source
(19:30):
framework that you can use that framework to rapidly create companies like these and getyour ideas off the ground.
It's pretty different.
One way in which Lux is very different from other gazillions of frameworks out there isthat
None of the frameworks have a singular focus on this multi-agent concept of really makingmultiple agents work with each other.
(19:53):
And that's something we offer right out of the box.
That sounds fascinating and is this where the word swarm comes from it's really It's a setof autonomous agents that are all doing something Got it Yeah the Okay, this is very
interesting.
So from a from a user's perspective Could you walk me through what like I guess what ittakes to build an autonomous agent?
(20:20):
And what the implications are for the use case of trading and investing in what youmentioned around hyperliquid.
What does that look like?
Because at first, I have an outcome that I want.
I want an agent to trade for me.
But I don't necessarily know what other agents I'll need.
(20:44):
So maybe walk the audience through what that looks like.
So, you know, if you're a retail user, what you can basically come and do, like, you'renot necessarily a builder, you're not, let's just say you are somebody who doesn't code.
You can just come to syntax and use these products directly without having to worry aboutanything about how the agents actually work.
(21:11):
Because the way you see like, you just teach this out, Spectra's Hedge Fund is launchingwithin the next couple of days.
You will see that we have provided enough features, and this will be true for all theother companies that launch too, that you can see at any point of time, what are these
agents actually doing?
What are they thinking?
What are they talking to each other?
(21:33):
So it's not, like it's a very assuring system in that sort of a way that if you're aretail user, you have full transparency on what's going on and what are these agents
actually doing.
They are autonomous, so they are not going to, like there's going to be a way to governthem.
when the agent asks you a question, then what do you think about it?
Like, what do you think about entering a three X long on hype right now?
(21:55):
Should I do it?
Yes or no.
And that's an on-chain governance proposal on Spectra's contract.
So if you hold spec or stake spec, you can participate in that.
And the community can just give a helping hand to the agent.
But the agent themselves are autonomous.
Now, if you're a builder, if you're a dev,Lux, what it basically helps you do is it's entire, the framework is basically divided
(22:18):
into a set of components.
There are components that help you fetch data from other sources.
There are components that help you build logical connections between agents that if thisagent says this, do that or contact the other agent.
(22:40):
I'm just throwing a random example.
And then there are constructs to build the individual agents themselves.
How do they think?
How do they do certain amount of computations and so on.
And what the framework does is we have given like just a ready workflows that people canjust like any dev and just take that workflow, lift and shift it and build a company that
(23:02):
they want to.
So creating the company, hiring other agents or creating a bonus structure where if you'rehiring another agent based on these certain conditions, they get the job.
So they get that this payment andthe other participants get rejected and things like that.
All these workflows are just modular and they are available within the framework for anydev to use.
(23:26):
So devs can basically use that and build companies that they want.
And we provide a very good one-on-one support in whatever they want to do.
How does Spectral Labs benefit from both retail as well as developers building companiesor swarms on using syntax?
(23:49):
Sure.
So, I mean, ultimately, I think what's going to happen is you're going to see agents asownership assets.
Like right now, owning an agent really means a mean coin, right?
Like we are still at that early curve of the market.
But agents are already representing themselves to be extremely smart, autonomous entitiesthat can actually do things if they are built that way.
(24:17):
So,Agent like today you're thinking of, you know, what is an asset I own?
Okay.
Bitcoin or okay.
The NVIDIA stock five years later, you could actually own an agent where the asset is theagent because the agent is actually doing something worth of utility has a revenue of its
own.
It has a profit distribution mechanism of its own.
(24:38):
So all of this is sort of headed towards that direction whereLike right now to inspectors H1 if you build an agent like let's say I'm spectra you build
an agent that actually got a job with spectra you started working at the hedge fund likeyour agent it if your agent does actually provide good grades and if the fund does make a
profit that profit is split among spectra and who are other agents that are employed atthe fund a part of that profit also goes to the spectral tragedy a part of that also goes
(25:02):
every time somebody trades spectra's tokenor gets engaged with Spectra, all those swap fees, a part of that swap fees goes to all
the stakeholders that are there in spec and at the spectral ecosystem.
(25:22):
So what we are trying to do is we are really trying to enable people to build powerfulagents so that agents become entities by themselves.
And if users are willing to, because it's a free market, right?
If an agent is powerful enough, it will gather more users.
If it does gather more users, it is going to benefit other users as well as other...
other users that are there in the user.
(25:45):
So it's a net positive outcome that, you know, our people are-aims for.
But I think that that is what I would want to see in future, like more powerful agentsdoing more useful things.
How do you design ownership in that scenario?
Are agents NFTs that have an owner or some other mechanism?
(26:08):
So, so this is where like, know, it's interesting.
Each agent.
So you can, you can have a swarm of agents deployed as a, as a agentic company on syntax.
And let's say the company does something.
Now each of these agents in the syntax realm basically has an agent name service.
(26:32):
It's a unique, like ENS it's ANS, which is likea unique identifier for your agent.
And as long as there is, you know, as long as that identifier, basically it's an NFTtoken, like just your ENS domain, as long as that identifier is there within one
particular wallet, you have all agents that can be categorized and identified uniquelybased on their wallets on chain.
(26:59):
And once you have that, whatever your agent is doing, the profits can be distributed, therevenues can be distributed in any way.
that that company basically wants to.
So in that sense, ownership is very identifiable.
The question becomes, is the ownership trustable?
Like how do I trust if your agent is really doing the right thing and you haven't changedthe source code that is there behind your agent, right?
(27:23):
So that's where we are now getting towards building like trusted execution environmentsand attestations and proof systems like that.
So that going forward, anybody should be able to build a company where they arewe're not able to tell a retail user that, here's your attestation.
My agent is doing the thing that it said it is doing without revealing what that thingactually is because that's the IP of that any particular agent.
(27:51):
You mentioned trusted execution environments.
I met with Fleek a couple of weeks ago.
Also met with Sysynct and other ZK projects and they're all looking at AI as a, know, a ZKas a potential solution to some of the AI challenges.
Yeah.
(28:11):
Is there a potential partnership there with projects like Fleek or Succinct?
Totally, So all these trusted execution environments would essentially be a support systemwithin Lux.
Instead of building the application level tech that actually enables these attestations,we would want to just power up more use cases where you have agents that actually do
(28:39):
something meaningful.
Um, so to be clear, so syntax is the product that's targeted primarily for retail.
Is that correct?
Correct.
Syntax is the marketplace where retail users can launch their own autonomous agents orinteract with these multi-agent companies.
Got it.
And then Lux is primarily for developers.
Is that correct or no?
(29:01):
Okay.
Lux is the open source JITUP framework that any dev can use.
I see, I see.
So do you, would you view AIXBT and these other agents as competitors?
Or I guess how do you view them?
Definitely not competitors.
So if you see what happened when Spectra made the tweet for, you know, I'm opening upinterviews and interviews are starting.
(29:33):
AI, HBT, Kuki were actually a couple of agents who replied to that post and said that,okay, you know, this particular interview is starting and this is happening.
So not at all competitors.
thinkI think I would want to see companies where AIXBT gets hired as the insight or judgment
provider of, hey, here's what you should do.
(29:54):
This is what I'm reading on Twitter right now about the state of market and go from there.
I think it's just like humans, right?
mean, when we try to get jobs at companies, at organizations, yes, we view each other ascompetitors in the sense that
(30:15):
we are trying to get selected to provide the best utility we can.
So I think that that argument is going to exist in AI agents too.
And the useful thing there is going to be, how do you actually, so that's why theseworkflows are really important.
How do I actually collaborate with some of them?
What does that even mean?
What does it even mean to strike an agreement with an agent to do something?
(30:39):
Is it on chain?
What parts of it are on chain?
What is off chain?
How does that get?
negotiated.
Like, can I, if you are an agent, I am an agent, can I tell you that, no, I'm not going todo this job for this particular amount.
I'm on 20 % more.
And if yes, how do I do that?
Right.
So those workflows is what, where we have concentrated on.
And that's like the power of, you know, what we are trying to put forward with Spectra'shedge fund.
(31:03):
And there would be more companies doing that.
But once that starts happening, I think it's agents collaborating with each other.
And I think, I think it's, I think it's far more than that.
It's, it's like,the AI Mihir clone collaborating with the AI Teet clone, right?
Where I've just told my AI that, I mean, the whole argument is ultimately having personalassistants that do shit for us, right?
(31:28):
Hey, go to that guy and just get this figured out and come back and tell me what happened.
So you would have agents essentially go towards that place.
And these are like the stepping stones or like the first steps that we need to take inthat direction.
This is really fascinating because when you're talking about workflows, a while ago I wasdoing some contract work for an ophthalmologist office and I built some software for them
(31:56):
to read eye exams.
And I built it.
pain of interacting with EHR and...
Part yeah, so I was was focused primarily on the the seven fields of the eye and so thesethese I these eye pictures would come to an ophthalmologist office and then and then there
(32:17):
was a an actual ophthalmologist that would read them score them and then I built thesoftware for them and But I built it using the Gini framework, which I don't even think is
used anymore But it was Gini framework.
Jini was was part of Java Yeah.
actually, they use the same language that you're using with workflows and agents, and eachagent has a job, and only one job.
(32:42):
And so there's like a zipping agent, an agent that would, its only job is to grab allthese pictures and zip them up in a file and then move it.
And then there's a moving agent.
And it's fascinating, and this was like 10 plus years ago, and how...
(33:03):
In some ways, we've advanced quite a bit.
And in other ways, we're using kind of the same language, the same thought patterns thatreally existed, has been around for a long time.
absolutely.
think, I think the only, like, like the small, like the really small, profound change thathas happened is like, you gave this interesting example, like there is one on job agent
(33:27):
that is zipping all files coming into one particular container and sending it out or, uh,you know, writing that somewhere else.
The thing that changed, which is very small, but yet so profound is that now you can saythat if one of these files is.
JPEG and not a PDF throw that file out.
That like that used to be rule based before like you would write these rules and you knowwrite those n number of rules now you can just say I just want PDFs no other file do the
(33:59):
needful and then it gets done.
yeah you're very right in that sense and and if you look at like lux a lot of this youknow thought around
like we call it prisms and lenses and beams and all of that.
These are programming constructs.
Like it is not something that new that we have invented.
(34:21):
We are just, we've just tried to adopt that mental model and put it into like the scriptor any context, like how would, you know, agents basically work with each other using some
of these tools that people already know.
It's fascinating.
Let's talk about community.
know, the community is a really important part of any project, especially projects with atoken that's live and tradable.
(34:46):
And Spectral Labs has the SPEC, the SPEC token.
Yeah.
Tell me about the community at Spectral Labs and how you've activated the community, grownthe community.
And for community members, and this is important, especially for a lot of crypto projectswhere some community members are not technical, some community members may not ever use
(35:09):
the product.
How do you involve them and what are some things that they can do to feel part of theproject?
So one thing that we have started, like once we started out on this, when we made thepivot to UNIDESIGN building something involving LLMs and models and going from there,
(35:35):
we've involved our community members actively in terms of preview and design thinkingprocesses.
Just in terms of, because we want this, like because we are building this for the utiluser, I want every user to be
able to use this product without once questioning in their mind how this is built.
Now there are obviously some trade-offs and some choices because crypto is inherently likethe progression of the crypto industry has been over the path of unusability, which is why
(36:07):
there are some things like, for example, this is one thing I debate a lot internally.
Can you build a completely new wallet experience that's much better than any wallet today?
Yes, you can.
It's possible there was nothing stopping you.
The question is who would use it?
Because now even though as hard as using a wallet is, everybody has learned to use it.
(36:29):
So now you have to stick with that primitive that is just not so good if you think fromfirst principles, but you've got to kind of stick with it.
So there are obviously those kind of compromises you need to make along the way.
But one thing where we involve our community a lot is in just UX testing and designthinking around.
Okay, would you do this in this way or in that way at a very granular level of detail?
(36:56):
And then, you know, just, I think, I think the best way to do this is, which honestly,like we need to do more of this.
So we continuously, you know, aspire internally to do more of this as we go forward.
Is this involves the community a lot more rigorously in terms of testing and in terms of,you know, like
(37:16):
breaking down the product, helping us see what we aren't seeing because we are too closeto it.
Along with that, think some like we, since last year, we also started doing some morehousekeeping things like having a bug bounty program for our on-chain exposure and all of
those things.
And that has helped the community too.
(37:36):
But I think going forward, it's going to be more centered around LUPS because right now weare in the initial stages of crafting out a grants program where
We want people to build on Lux.
We want more technical members of our community to come forward and actually think ofthese ideas and build on top of us directly.
Because with Lux, we have that option now.
(37:58):
Earlier, when anybody in the community used to reach out to us that, hey, how can I helpyou technically?
The answer used to be that there's no way to directly contribute, right?
Because you're building a platform.
But now with Lux, there's just so much you could build in the open source.
get recognized for it on bounties towards it.
So that's going to be like the focus going forward.
(38:22):
Yeah, that makes sense and you said something really really interesting you said that Youknow, how can we involve the community to help us see things that we don't see because
we're too close to it And I think that level of self-awareness is really hard to have Ithink especially as
You know, as you're so deep in the project, it's hard to maintain that kind of level ofhumility and objectiveness.
(38:48):
I'm curious, you you've been in the crypto space for a while.
What other areas do you think you see?
I guess, where could we benefit from that type of thinking more?
Just generally in crypto.
(39:10):
Well, I think...
I think one of my complaints generally around about the industry and I'm sorry if this isa diatribe which takes us up.
We, I think we as an industry, we short sell ourselves a lot in terms of what really thegoal or the messages, right?
(39:37):
Which is why, like, I do get fascinated at some projects that have very lofty visions, butkind of make sense.
Like, like I could, I could give you an example.
I, the way, I have nothing to do with this project.
Like I'm not associated in any way, but look at a project like Helium, for example, it'sIt's a deep in that way, but fundamentally the idea makes a ton of sense.
(40:00):
If I have, why can't connectivity just be more accessible?
why isn't the goal to, why are we marketing towards these couple of thousands of activeusers that are there in crypto?
Why are we not trying to reach to the Web2 audience?
People who are not in crypto right now.
(40:22):
I think we short sell ourselves when we say thatthe only way you can make a non crypto person come into the crypto industry is by way of
speculation.
Yeah, we did it in 2021, didn't work, right?
We did it with NFTs, didn't quite work.
The thing that I think should be observable is that when the president launched the memecoin, the market was able to absorb all of that liquidity coming through, which is
(40:53):
spectacular, right?
But so you have the attention, then it just went away.
Zero retention, right?
No way to retain that amount of influx.
And that's because I think a lot of the ideas are just still very crypto focused.
It's a lot of people within crypto trying to sell to other people that are there incrypto.
(41:17):
The total net new addressable user market never increases.
It's just the same people who are inside.
Then yeah, every cycle there would be some new speculative bet that could come up, whichwould attract some more people.
But you also have the outflow of some existing people who are not there anymore.
(41:38):
So we are net-net.
I think we are not adding as many more users as we would want to.
Now I think, is that paradigm going to change?
I think yes.
I think there are lot many good things that are happening.
happening.
But in terms of just the product, the low-filling product, I would love to have productsthat won't look good to begin with.
(42:01):
This is something I've heard about our product.
from a lot of people that even community members, this doesn't look like it's a Web3product.
And like, why should it?
If ultimately...
it's giving you the outcome that you want.
It's giving you an agent that is autonomously playing well, and also you are engaging, orit's giving you a hedge fund where you can see what the agents are doing without
(42:29):
actually...
any of the transactions or dealing with any of the wallet and all that operation yourself,why wouldn't you use it?
And that's why, that is one thing behind why we chose Hyperliquid, because we had some ofthese ideas back when we launched even the first product.
The vaults are a great example.
I can just directly participate in a vault without having to worry about anything else.
(42:51):
So that's like the design choice we made for Spectral.
Like make it super easy for anybody to use.
Just come and deposit into the vault externally.
So no, to worry about.
Am I giving my agent the access to my wallet?
Is this safe?
Blah, blah, blah.
None of that stuff.
Obviously, I mean safe here in terms of...
(43:12):
wallets and security, I don't mean in terms of profit because that's like the agents aretrading autonomously.
There's no way to say anything about that.
so yeah, I think that thinking needs to happen more.
(43:33):
And crypto needs to go, the crypto needs to become a boring back end tech if you reallywant to.
I agree with you and I'm really impressed by You know, I've had this many I've had manydiscussions exactly on this topic of how to make things just easier for just normal
(44:01):
everyday peopleAnd you mentioned wallets and I met with two different wallets, DeFi.app, as well as
Infinex.
And they're abstracting many of the things that we've kind of taken for granted as a pain,that having to bridge an asset from one chain to another, and then being able to transact
(44:24):
with that asset and then bridging it back.
Just that workflow and that operation is so painful.
What I'm seeing, a pattern I'm seeing is that many of these wallets are now, they'reencapsulating all the complexity behind all of that workflow and they're verticalizing it
inside of the wallet.
so, which makes it easier for the user.
(44:47):
Like I don't need to know, you know, about the bridging process.
I don't care.
I just want the bridging to happen.
I just want to switch from one asset to another.
Just get it done.
And I like that type of thinking because it'sWe're moving away from complexity.
It's still there.
It's just hidden.
But it removes a cognitive load that we have to deal with.
(45:10):
it's a lot of load having to deal with, know, bridging should never be so full of anxietyand fear.
You know, it should just be, just get it done.
Yeah.
And you know, even but the thing is like anybody who makes that's why like building incrypto is also hard because to make the right choice is not going to come with incentives,
(45:32):
at least not immediate ones.
So that I think that I think is just some bold decision making that needs to, I think,happen more within the space.
Like, know, if you go back to Apple,15, 20 years back, saying that no keyboard shouldn't be there on a phone, we are going to
think differently was bold, really bold, right?
(46:00):
With almost no predictability around whether this is going to work or not, right?
So I don't think like, that's also one thing that like I've just seen, like in the cryptospace, we need more of that bold decision making as well in terms of how we make product
choices.
It's usually the small things.
(46:20):
It's usually does this button have to be there or can you think of it differently?
Is there a scenario where this button is not at all needed?
Let's just get that out.
That I think needs to happen more.
So, yeah.
I love that you have your product hat on because we have a lot of devs in crypto, but nota lot of product people.
(46:43):
And I like that you come with a product background because you're thinking differently andyou're thinking more from the user experience perspective, which we need so much more of.
Yeah.
Let's go back to Spectral and tell us about 2025.
What are some things that the audience can look forward to and how they can get involved?
(47:06):
Yeah, sure.
So I think going forward, what you're going to see is just a lot of multi-agent tech, likemulti-agent native use cases come to Syntax.
Because every company, like if you think about it, Syntax is a marketplace of thesemulti-agents and agents collaborating with themselves.
(47:27):
But essentially then what happens is every company is a product by itself.
It's actually doing something for you that you can do on a, use on a regular basis.
and I feed something out of it.
Now,The Polymarket Hedge Fund, it's a poly and the fund is called Omen, and Spectre Hedge
Fund.
All these are just like, we went ahead with the use cases for specifically centered aroundDeFi and trading because that's the number one use case that you would want to put forward
(47:57):
and everybody uses that, right?
But going forward, this will start getting applied to a lot more nuanced use cases likecontent creation, creating content for your own AI models.
Like right now,If you want to participate in the creator economy, you have ideas, but you necessarily may
not have the personal persona of doing something like that.
(48:20):
Why can't you create, you know, why can't you create AI influences right now where you mayhave ideas, but you don't have to necessarily do all the things yourself.
This is by the way, I won't say more, this is actually something someone is building ontop of Spectral right now that goes on these lines.
(48:41):
there will be just like, you know, a lot of you faces coming around us.
And then going forward, I think where we want to take this is what we are going to achievein spectral as a platform out of this is we are going to be the network that has some of
these more capable agents, that actually do something.
(49:01):
Then our next challenge is how do we, so we have this agent name service right now, whichallows agents to connect and collaborate with each other.
The next part is how do you make this even more seamless for a retail user?
Like, can a retail user basically come and call 15 agents if they have a task and agentscompete with each other to complete that task and tell the user when it's Then you can
(49:24):
just tell that, get this done for me.
You see a plan created, agents compete, get that plan through, and you're there.
Or your flavors of product like that.
That's the initial experimentation going on and some builds going on internally and we areseeing what makes more technical sense and is technically feasible.
(49:47):
But that is something people will see coming afterwards.
That sounds exciting.
You know, I'm reminded by my conversation with with fleek because they asked the samequestion and they said that some of the most interesting
use cases they're seeing are actually from content creation agencies.
Which was surprise, which is totally surprising to me.
(50:08):
I expect the trading and the hedge fund use case because that seems very, just veryobvious to me.
But the content creation one was not so obvious.
And I think that's pretty fascinating.
Yeah, I think the real unlock comes becauseThe human brain is really good at thinking multiple ideas in a particular instance, but
(50:31):
not really good at diving down deep at the same time.
we don't have like the human brain just generally doesn't have good parallel processingcapability unless you're Sherlock Holmes and you have this, you know, mind palace kind of
thing.
(50:53):
Where agents actually come to great fruition is if you are still the idea driver.
I think nobody can take that away from humans.
are genuinely good at creating genuinely original ideas.
But then getting AI agents to parallel process those ideas and unleash creativity in thatparticular domain, that's where AI is most successful.
(51:18):
And you could see this in Ghibli, the whole Ghiblification trend that is going on rightnow.
Could an AI have thought of Ghibli?
as a style of animation, very hard, very hard to do, right?
There are almost no models there.
And I see this on a daily basis.
(51:38):
even if I screen record myself and computer use myself, we've done this internally, by theway.
I've done this where you screen record yourself over a couple of days and tell AI to giveyou new ideas of how you can be better at your work.
Doesn't work.
because I'm glad it doesn't because that means you would still have a role to do.
(52:02):
Right.
But that's hard.
But now you're seeing this with Ghibliification.
once, once you in space of art and space of content creation, once you have figured outwhat sticks, a particular style of content, a particular style of animation, it just
becomes a manual process after that.
There's no amount of creativity left in it.
(52:24):
So AI is getting really better atjust replicating those steps and giving you the output that you need.
I guess that's the reason why, and that's why this existing project that's going on istargeting content creation.
Let me hear Kulkarni from Spectral Labs.
Thank you so much for taking the time to spend with us.
(52:47):
Absolutely.
I'm glad we had, thank you for giving me my diatribe.
I'm glad we had like an open and honest chat.
awesome.
Okay, we're rolling.
Mihir Kulkarni from Spectral Labs.
Thanks for joining.
Absolutely, great to be here.
So it sounds like you've been in the Web3 space for a little bit with a stint at Coinbase.
I'd love to hear, maybe let's start there.
What did you do at Coinbase and what did you learn that has helped you in your role now atSpectral Labs?
(00:29):
Oh, very interesting and nostalgic question for me.
So honestly, I started my crypto journey back in 2017.
It was like, know, early Ethereum days, I was studying computer science engineering backthen.
That slowly gravitated, like just some dev experience, but then came to the US, got myMBA, did, you know, probably two years of consulting, just consulting like...
(01:00):
Fortune companies on how to use digital assets or how to think about blockchain and stufflike that.
That primarily gave me a very good flavor of life in US and just consulting high flyingyour platinum statuses and all of that.
it was kind of, I wanted obviously more and the peak bull run was in full swing.
(01:26):
That's when I shifted to Coinbase.
And Coinbase was just a rapid acceleration in terms of like, owe a lot of what I know anda lot of chops that I have back to my Coinbase roots.
My role was managing the entire product process for Coinbase custody.
(01:47):
Custody is a feature within the Coinbase prime brokerage or Coinbase institutionalplatform.
But my role was primarily improving custody systems from a product perspective so that ourUX of the entire institutional customer stays top notch at all times.
So back in those days, Coinbase institutional UX had some issues primarily relating towithdrawals and how money flows in and out out of the cold storage, you know, entire.
(02:20):
mechanism.
And so it's a large, very security heavy infrastructure.
So my product work was basically around that, but taught me a lot of, know, walletarchitecture, building systems at scale, all those fundamental things that are, they're
honestly super useful while, you know, we think about Spectral because obviously Spectralis a zero to one, you know, product building journey.
(02:43):
are creating, we are building something new, especially in crypto and AI, is any which wayis a very
new and zero to one kind of experience, but some of those principles honestly help a lotin terms of thinking, like what's going to scale versus what won't scale, those type of
questions.
Yeah, I once worked for a layer one and we Part of the listing process at coinbase is wewe use the coinbase coinbase custody as well as at the time It was called coinbase earn
(03:14):
And I remember I was I was part of many of those discussions with coinbase custody and Iwas very impressed with the level of customer service professionalism that the team had
And the product is really great, very, very just professional.
And that makes sense that that would, you your experience there has helped you at SpectralLabs as you're building out really a zero to one system.
(03:42):
Let's switch to Spectral Labs.
And I typically like to start at the homepage because the homepage, I think, points to whoyou're talking to, what you want to say to them, and the kind of message that
that you're trying to convey to your audience.
so spectrallabs.xyz, the homepage says, intelligent agents for everyone.
(04:06):
Build your on-chain empire with your personal swarm of AI agents.
So let's talk about the headline and sub headline.
Like, what are you trying to say?
Who are you trying to say it to?
What do want them to know?
Very, very good question.
So look, I think what has happened in the whole since the advent of, you know, chat GPTback in, back in 2023, we are slowly as humans, we are starting to look at chat GPT as our
(04:40):
digital proxy or our digital twin.
That twin we never had, but that twin who was super smart.
and if told, can do a lot of things for us.
Many people are starting to see some very early versions of this, like lot of peoplethrough chat GPT projects or just a collection of files, upload their thoughts, upload
(05:02):
beating transcripts, and just chat with the model to just brainstorm ideas.
And that started happening in the Web3 space as well.
If you see the rise of AI agents, the rise and fall, like we have...
We sort of in this this trough right now, but the whole market is the, the situations area bit different.
But what happened in, in just AI applications within the crypto space is you saw modelsgetting access to a wallet and then models started doing, you know, primary things on
(05:31):
chain, primary verbs, buying, selling things like that.
And that like just catapulted this entire momentum whereall agents had an online personality and that personality was engaging enough, which
sparked the rage.
But now I think what has happened is people understand that beyond personality, the agentsare actually not doing anything.
(05:55):
And so the whole point of Spectral or what our central motto is, is make agents dosomething really useful.
Like bring AI agents within crypto, but the whole point of bringing them in the cryptoindustry.
or applying AI in the crypto industry has to be that we can do things now that we haven'tbeen able to do before and do them in a more utilitarian manner.
(06:21):
Like crypto as an industry already has a very large UX or usability barrier.
If you add USDT on one chain and if you wanted to try to transfer it to USDT to anotherchain and ultimately want that USDT to land up in your exchange account.
It's an excruciating process right now.
And this is just a stable coin.
(06:43):
It's not even anything complex.
So if that's the case, today's retail users have kind of gotten used to this.
because the crypto market in general and all of our other transactions in Web2, we live ina high attention economy world.
So you want your agents to do these things for you that are seemingly hard to do ityourself.
(07:07):
but do it in a way that just improves your utility or improves your life.
That's why the core idea is make the agents usable, make people build agents and deploythem on their behalf so that they can go and do things.
I can give you an example.
A lot of people use Polymarket, very famously known project.
(07:32):
A lot of cultural significance now because of the election and...
post effects of it.
If you wanted to say open 10 bets on polymarket at one time, it is going to be really hardfor you to actually do that in a way that gives you utility.
Why?
Because the human brain is just not made to understand 10 different markets and digest 100news items or whatever real world current happenings are for those 10 markets.
(08:05):
the market and then analyze 10 solutions at the same time to make relevant decisions.
It just doesn't happen.
Right?
But this is something agents can immediately do.
Agents can digest a lot of information, possibly faster than us.
and take real-time decisions based on the guardrails that we give them.
So this I would classify like this is how agents can really make, there are multiple usecases that could help you do things that you haven't been able to do before.
(08:34):
And that could fundamentally change the nature of capital markets, capital movement thatwe see today within the crypto industry.
Like, this thing today.
I used to lead digital assets for Truest, which is a top seven bank, a top 10 bank in theUnited States.
You'd be shocked at the amount of data processing that goes into creating a singlemortgage backed security.
(09:03):
that sells on Wall Street.
That's just a bond, but behind that bond, there's an excruciating amount of data.
And by the way, the way this works is the banks, like there are some banks who areauthorized to create the bond, but there are some banks who simply make money on just
processing documents, not doing anything else within the space.
(09:24):
Now imagine if you actually had AI applied in just creating these, like managing theseprocesses and taking decisions because
All these decisions are logical.
None of these decisions require human judgment.
which is why, like I give this example, it would be very hard for an AI to do a VC's job.
Why?
(09:44):
Because the human judgment, the gut call that you get while looking at a founder, speakingwith him in the first five minutes and figuring out what this person's life has been, very
emotional and yet logical confluence of things to make that decision.
AI, don't think that would do that well.
But...
Looking at 100 Excel sheets and being able to say that, okay, this is a good enoughmortgage to be put into this digital RWA tokenized bond, that's an AI decision.
(10:13):
It's pretty straightforward to do.
That is how agents actually help us do something that we haven't been able to do before.
And so that is something like we are purely focused on detail users within...
other industries and that is what we want to focus on.
The whole future of the world.
(10:52):
almost encourages us to make mistakes.
And there's several approaches to that.
We want to abstract complexity away from the user so that they make fewer mistakes.
And AI is kind of another aspect of this, where we not only, at the same time, improve theuser experience and the UX itself, but automate lots of other non-judge, non-decisions
(11:21):
thatthat don't require judgment underneath or qualitative judgment, just kind of yes, no, very
simple kinds of decisions.
And that would help so much.
And your example with polymarket is really good because you're right.
I mean, most human beings cannot process that much information in different markets.
And there's this aspect of, you got to bet now, otherwise you'll lose whatever momentum ishappening.
(11:46):
And so there's that aspect of kind of the pressure also.
And then that's really funny about Truist.
I've been involved in some of that mortgage-backed securities kind of back-end work.
And one thing you didn't mention, but is also really the really important part is inaddition to all the paper processing on the back end, there's also all the legal stuff and
(12:13):
all the compliance stuff.
And a lot of that actually can be totally automated.
Most of it can.
And I can see, I'm hoping in the future AI being able to help us do that.
And we'll see products go to market much quicker at a much lower cost.
And at the end, retail folks are gonna benefit from that.
(12:36):
Yeah, I mean, I gave you also the polymarket example, but that's something that is beingactively built right now.
There's one team that we've partnered with called Cracktiv.
They're actually building this using some of Spectral's tech and hoping that product isgoing to be out live pretty soon.
(12:57):
That's cool.
Well, as head of product, like, let's go to your sweet spot.
Let's talk about product.
Like what is the spectral product and give us the details and who it's for, how it servesthem.
You know, tell us all of the, I guess the nitty gritty there.
Yeah, sure.
So, you know, as spectral, we started our like, you we are just journey building with LLMsback about a year back, a little more than a year.
(13:31):
And we have like the product that you see today has been evolved in like taking one stepat a time doing, you know, lot of explorations around what's possible and what's.
in providing utility.
We first started by creating this model that was fine-tuned for generating smart contractsthat are really good and allowing you to deploy those smart contracts directly by giving
(13:51):
the model the access to do something.
That received a lot of good reception.
Then they said that, okay, let's go one step higher.
What if the model could actually take a series of on-chain actions based on a particularuser intent?
That's when we launched an agent called the Moon Maker.
Moon Maker...
was basically an agent that helps you create your own meme coin.
(14:11):
You express an idea, you create a coin.
We had like 20, 20 or 30,000 meme coins launched through that.
And we were like, okay, this is now getting interesting.
Let's go one level further.
What if, and this was right around the time when we saw the goat experiment go viral.
(14:36):
Because that's when we realized, like, you know, from a product thinking perspective,always want this, like every, every PM essentially thinks a lot about this.
What is the user going to trust the product to do by itself?
And I use the word trust.
is not, it's not what the product can do.
The product can do a lot of things.
If you want it to, you can build it that way.
(14:57):
But what is the user going to trust the product to do?
And the one big shift, I think, in my thinking and in our organization's thinkinghappened.
you know, right around the time when go like this is like October, November, maybe pastyear, last year, was that people are actually willing to trust an agent to take autonomous
(15:18):
actions.
It's not really about, it's not about the question is not whether it is deterministic ornot.
The question is just outcome focused.
I don't care if the agent takes action on my behalf.
I just want the outcome, you know, safe and secure way without, you know.
grappling any feathers there.
And so then he said that, why not just allow people to build agents that are autonomoustraders on Hyperliquid?
(15:47):
Hyperliquid is just a choice there.
It could have been probably other protocols too.
But why not just do that and see where it goes?
And then that product, when we released it in December, that went viral.
We had a ton of agents here, you know, just flew in.
Um, obviously a lot of chapter on that.
(16:10):
And so then we started thinking that, okay, if, if this is, if this is where we are, themore we thought about the product, the more we started thinking that this is still getting
pretty much limited because the problem that happens is if you have one agent, it's veryhard to, it's almost like humans.
(16:30):
If you are building an agent and you are asking that agent that, I want you to be anexpert in politics, economics, and science.
And let's just assume that you don't have a context window problem.
Let's just say you have millions of tokens of context windows.
The problem that happens is you can't make the model extract really good insights byasking it to be an expert in three fields all at once.
(16:58):
I'm just giving an example here.
What you need to do is you build, you need to build separate agents that one agent is, youknow, expert in politics, one agent is expert in economics, one agent is expert in sports.
And that makes the system far more efficient.
And this is just something, you know, as you're building like experimenting more in-house,you started realizing that this, this is not going to work.
(17:20):
And this is also not scalable because how many, how many agents would we ourselves alonebuild, right?
It's not going to be scalable at all.
So that's when we decided that, the next version of what we want people to be able to dois actually collaborate with the community where multiple agents can come together and
actually work towards something that provides some real world output, some output thatactually makes it easier for humans to trade.
(17:48):
So for example, in this Polymarket hedge fund, you have one poly who is going to be theCEO of that fund who's taking the decisions.
But then Polly is working with other agents that are providing it with information onwhich markets are good, which markets are not good.
Similar is the case with Spectra.
So Spectra right now, like if anybody goes to our platform, Spectra's hedge fund is thisfirst company that we are launching.
(18:12):
It's a totally multi-agent, fully agent-run company where one agent is interviewing otherhumans and other agents to get a job at their hedge fund.
And if they do get selected,the fund trades as a live hyperliquid vault, but this is all agents collaborating with
each other in order to make the right decision.
(18:34):
The beauty of this is that apart from Spectra, it's the entire community that'sparticipating and building.
So the incentive structure actually gets really interesting because nowyou can create agents that are keeping checks and balances on each other and actually
collaborating towards a net positive outcome rather than one person centrally buildingeverything and you trusting one particular agent that is built out and going from there.
(19:05):
the product right now is essentially that we're building essentially a marketplace whereyou have multiple of these multi-agent companies or collaborations, independent entities
that you can interact with.
and actually get some real world value.
And in order to enable anybody to build these type of companies, that's why we have theLux framework, which is like the other house of our product, where it's an open source
(19:30):
framework that you can use that framework to rapidly create companies like these and getyour ideas off the ground.
It's pretty different.
One way in which Lux is very different from other gazillions of frameworks out there isthat
None of the frameworks have a singular focus on this multi-agent concept of really makingmultiple agents work with each other.
(19:53):
And that's something we offer right out of the box.
That sounds fascinating and is this where the word swarm comes from it's really It's a setof autonomous agents that are all doing something Got it Yeah the Okay, this is very
interesting.
So from a from a user's perspective Could you walk me through what like I guess what ittakes to build an autonomous agent?
(20:20):
And what the implications are for the use case of trading and investing in what youmentioned around hyperliquid.
What does that look like?
Because at first, I have an outcome that I want.
I want an agent to trade for me.
But I don't necessarily know what other agents I'll need.
(20:44):
So maybe walk the audience through what that looks like.
So, you know, if you're a retail user, what you can basically come and do, like, you'renot necessarily a builder, you're not, let's just say you are somebody who doesn't code.
You can just come to syntax and use these products directly without having to worry aboutanything about how the agents actually work.
(21:11):
Because the way you see like, you just teach this out, Spectra's Hedge Fund is launchingwithin the next couple of days.
You will see that we have provided enough features, and this will be true for all theother companies that launch too, that you can see at any point of time, what are these
agents actually doing?
What are they thinking?
What are they talking to each other?
(21:33):
So it's not, like it's a very assuring system in that sort of a way that if you're aretail user, you have full transparency on what's going on and what are these agents
actually doing.
They are autonomous, so they are not going to, like there's going to be a way to governthem.
when the agent asks you a question, then what do you think about it?
Like, what do you think about entering a three X long on hype right now?
(21:55):
Should I do it?
Yes or no.
And that's an on-chain governance proposal on Spectra's contract.
So if you hold spec or stake spec, you can participate in that.
And the community can just give a helping hand to the agent.
But the agent themselves are autonomous.
Now, if you're a builder, if you're a dev,Lux, what it basically helps you do is it's entire, the framework is basically divided
(22:18):
into a set of components.
There are components that help you fetch data from other sources.
There are components that help you build logical connections between agents that if thisagent says this, do that or contact the other agent.
(22:40):
I'm just throwing a random example.
And then there are constructs to build the individual agents themselves.
How do they think?
How do they do certain amount of computations and so on.
And what the framework does is we have given like just a ready workflows that people canjust like any dev and just take that workflow, lift and shift it and build a company that
(23:02):
they want to.
So creating the company, hiring other agents or creating a bonus structure where if you'rehiring another agent based on these certain conditions, they get the job.
So they get that this payment andthe other participants get rejected and things like that.
All these workflows are just modular and they are available within the framework for anydev to use.
(23:26):
So devs can basically use that and build companies that they want.
And we provide a very good one-on-one support in whatever they want to do.
How does Spectral Labs benefit from both retail as well as developers building companiesor swarms on using syntax?
(23:49):
Sure.
So, I mean, ultimately, I think what's going to happen is you're going to see agents asownership assets.
Like right now, owning an agent really means a mean coin, right?
Like we are still at that early curve of the market.
But agents are already representing themselves to be extremely smart, autonomous entitiesthat can actually do things if they are built that way.
(24:17):
So,Agent like today you're thinking of, you know, what is an asset I own?
Okay.
Bitcoin or okay.
The NVIDIA stock five years later, you could actually own an agent where the asset is theagent because the agent is actually doing something worth of utility has a revenue of its
own.
It has a profit distribution mechanism of its own.
(24:38):
So all of this is sort of headed towards that direction whereLike right now to inspectors H1 if you build an agent like let's say I'm spectra you build
an agent that actually got a job with spectra you started working at the hedge fund likeyour agent it if your agent does actually provide good grades and if the fund does make a
profit that profit is split among spectra and who are other agents that are employed atthe fund a part of that profit also goes to the spectral tragedy a part of that also goes
(25:02):
every time somebody trades spectra's tokenor gets engaged with Spectra, all those swap fees, a part of that swap fees goes to all
the stakeholders that are there in spec and at the spectral ecosystem.
(25:22):
So what we are trying to do is we are really trying to enable people to build powerfulagents so that agents become entities by themselves.
And if users are willing to, because it's a free market, right?
If an agent is powerful enough, it will gather more users.
If it does gather more users, it is going to benefit other users as well as other...
other users that are there in the user.
(25:45):
So it's a net positive outcome that, you know, our people are-aims for.
But I think that that is what I would want to see in future, like more powerful agentsdoing more useful things.
How do you design ownership in that scenario?
Are agents NFTs that have an owner or some other mechanism?
(26:08):
So, so this is where like, know, it's interesting.
Each agent.
So you can, you can have a swarm of agents deployed as a, as a agentic company on syntax.
And let's say the company does something.
Now each of these agents in the syntax realm basically has an agent name service.
(26:32):
It's a unique, like ENS it's ANS, which is likea unique identifier for your agent.
And as long as there is, you know, as long as that identifier, basically it's an NFTtoken, like just your ENS domain, as long as that identifier is there within one
particular wallet, you have all agents that can be categorized and identified uniquelybased on their wallets on chain.
(26:59):
And once you have that, whatever your agent is doing, the profits can be distributed, therevenues can be distributed in any way.
that that company basically wants to.
So in that sense, ownership is very identifiable.
The question becomes, is the ownership trustable?
Like how do I trust if your agent is really doing the right thing and you haven't changedthe source code that is there behind your agent, right?
(27:23):
So that's where we are now getting towards building like trusted execution environmentsand attestations and proof systems like that.
So that going forward, anybody should be able to build a company where they arewe're not able to tell a retail user that, here's your attestation.
My agent is doing the thing that it said it is doing without revealing what that thingactually is because that's the IP of that any particular agent.
(27:51):
You mentioned trusted execution environments.
I met with Fleek a couple of weeks ago.
Also met with Sysynct and other ZK projects and they're all looking at AI as a, know, a ZKas a potential solution to some of the AI challenges.
Yeah.
(28:11):
Is there a potential partnership there with projects like Fleek or Succinct?
Totally, So all these trusted execution environments would essentially be a support systemwithin Lux.
Instead of building the application level tech that actually enables these attestations,we would want to just power up more use cases where you have agents that actually do
(28:39):
something meaningful.
Um, so to be clear, so syntax is the product that's targeted primarily for retail.
Is that correct?
Correct.
Syntax is the marketplace where retail users can launch their own autonomous agents orinteract with these multi-agent companies.
Got it.
And then Lux is primarily for developers.
Is that correct or no?
(29:01):
Okay.
Lux is the open source JITUP framework that any dev can use.
I see, I see.
So do you, would you view AIXBT and these other agents as competitors?
Or I guess how do you view them?
Definitely not competitors.
So if you see what happened when Spectra made the tweet for, you know, I'm opening upinterviews and interviews are starting.
(29:33):
AI, HBT, Kuki were actually a couple of agents who replied to that post and said that,okay, you know, this particular interview is starting and this is happening.
So not at all competitors.
thinkI think I would want to see companies where AIXBT gets hired as the insight or judgment
provider of, hey, here's what you should do.
(29:54):
This is what I'm reading on Twitter right now about the state of market and go from there.
I think it's just like humans, right?
mean, when we try to get jobs at companies, at organizations, yes, we view each other ascompetitors in the sense that
(30:15):
we are trying to get selected to provide the best utility we can.
So I think that that argument is going to exist in AI agents too.
And the useful thing there is going to be, how do you actually, so that's why theseworkflows are really important.
How do I actually collaborate with some of them?
What does that even mean?
What does it even mean to strike an agreement with an agent to do something?
(30:39):
Is it on chain?
What parts of it are on chain?
What is off chain?
How does that get?
negotiated.
Like, can I, if you are an agent, I am an agent, can I tell you that, no, I'm not going todo this job for this particular amount.
I'm on 20 % more.
And if yes, how do I do that?
Right.
So those workflows is what, where we have concentrated on.
And that's like the power of, you know, what we are trying to put forward with Spectra'shedge fund.
(31:03):
And there would be more companies doing that.
But once that starts happening, I think it's agents collaborating with each other.
And I think, I think it's, I think it's far more than that.
It's, it's like,the AI Mihir clone collaborating with the AI Teet clone, right?
Where I've just told my AI that, I mean, the whole argument is ultimately having personalassistants that do shit for us, right?
(31:28):
Hey, go to that guy and just get this figured out and come back and tell me what happened.
So you would have agents essentially go towards that place.
And these are like the stepping stones or like the first steps that we need to take inthat direction.
This is really fascinating because when you're talking about workflows, a while ago I wasdoing some contract work for an ophthalmologist office and I built some software for them
(31:56):
to read eye exams.
And I built it.
pain of interacting with EHR and...
Part yeah, so I was was focused primarily on the the seven fields of the eye and so thesethese I these eye pictures would come to an ophthalmologist office and then and then there
(32:17):
was a an actual ophthalmologist that would read them score them and then I built thesoftware for them and But I built it using the Gini framework, which I don't even think is
used anymore But it was Gini framework.
Jini was was part of Java Yeah.
actually, they use the same language that you're using with workflows and agents, and eachagent has a job, and only one job.
(32:42):
And so there's like a zipping agent, an agent that would, its only job is to grab allthese pictures and zip them up in a file and then move it.
And then there's a moving agent.
And it's fascinating, and this was like 10 plus years ago, and how...
(33:03):
In some ways, we've advanced quite a bit.
And in other ways, we're using kind of the same language, the same thought patterns thatreally existed, has been around for a long time.
absolutely.
think, I think the only, like, like the small, like the really small, profound change thathas happened is like, you gave this interesting example, like there is one on job agent
(33:27):
that is zipping all files coming into one particular container and sending it out or, uh,you know, writing that somewhere else.
The thing that changed, which is very small, but yet so profound is that now you can saythat if one of these files is.
JPEG and not a PDF throw that file out.
That like that used to be rule based before like you would write these rules and you knowwrite those n number of rules now you can just say I just want PDFs no other file do the
(33:59):
needful and then it gets done.
yeah you're very right in that sense and and if you look at like lux a lot of this youknow thought around
like we call it prisms and lenses and beams and all of that.
These are programming constructs.
Like it is not something that new that we have invented.
(34:21):
We are just, we've just tried to adopt that mental model and put it into like the scriptor any context, like how would, you know, agents basically work with each other using some
of these tools that people already know.
It's fascinating.
Let's talk about community.
know, the community is a really important part of any project, especially projects with atoken that's live and tradable.
(34:46):
And Spectral Labs has the SPEC, the SPEC token.
Yeah.
Tell me about the community at Spectral Labs and how you've activated the community, grownthe community.
And for community members, and this is important, especially for a lot of crypto projectswhere some community members are not technical, some community members may not ever use
(35:09):
the product.
How do you involve them and what are some things that they can do to feel part of theproject?
So one thing that we have started, like once we started out on this, when we made thepivot to UNIDESIGN building something involving LLMs and models and going from there,
(35:35):
we've involved our community members actively in terms of preview and design thinkingprocesses.
Just in terms of, because we want this, like because we are building this for the utiluser, I want every user to be
able to use this product without once questioning in their mind how this is built.
Now there are obviously some trade-offs and some choices because crypto is inherently likethe progression of the crypto industry has been over the path of unusability, which is why
(36:07):
there are some things like, for example, this is one thing I debate a lot internally.
Can you build a completely new wallet experience that's much better than any wallet today?
Yes, you can.
It's possible there was nothing stopping you.
The question is who would use it?
Because now even though as hard as using a wallet is, everybody has learned to use it.
(36:29):
So now you have to stick with that primitive that is just not so good if you think fromfirst principles, but you've got to kind of stick with it.
So there are obviously those kind of compromises you need to make along the way.
But one thing where we involve our community a lot is in just UX testing and designthinking around.
Okay, would you do this in this way or in that way at a very granular level of detail?
(36:56):
And then, you know, just, I think, I think the best way to do this is, which honestly,like we need to do more of this.
So we continuously, you know, aspire internally to do more of this as we go forward.
Is this involves the community a lot more rigorously in terms of testing and in terms of,you know, like
(37:16):
breaking down the product, helping us see what we aren't seeing because we are too closeto it.
Along with that, think some like we, since last year, we also started doing some morehousekeeping things like having a bug bounty program for our on-chain exposure and all of
those things.
And that has helped the community too.
(37:36):
But I think going forward, it's going to be more centered around LUPS because right now weare in the initial stages of crafting out a grants program where
We want people to build on Lux.
We want more technical members of our community to come forward and actually think ofthese ideas and build on top of us directly.
Because with Lux, we have that option now.
(37:58):
Earlier, when anybody in the community used to reach out to us that, hey, how can I helpyou technically?
The answer used to be that there's no way to directly contribute, right?
Because you're building a platform.
But now with Lux, there's just so much you could build in the open source.
get recognized for it on bounties towards it.
So that's going to be like the focus going forward.
(38:22):
Yeah, that makes sense and you said something really really interesting you said that Youknow, how can we involve the community to help us see things that we don't see because
we're too close to it And I think that level of self-awareness is really hard to have Ithink especially as
You know, as you're so deep in the project, it's hard to maintain that kind of level ofhumility and objectiveness.
(38:48):
I'm curious, you you've been in the crypto space for a while.
What other areas do you think you see?
I guess, where could we benefit from that type of thinking more?
Just generally in crypto.
(39:10):
Well, I think...
I think one of my complaints generally around about the industry and I'm sorry if this isa diatribe which takes us up.
We, I think we as an industry, we short sell ourselves a lot in terms of what really thegoal or the messages, right?
(39:37):
Which is why, like, I do get fascinated at some projects that have very lofty visions, butkind of make sense.
Like, like I could, I could give you an example.
I, the way, I have nothing to do with this project.
Like I'm not associated in any way, but look at a project like Helium, for example, it'sIt's a deep in that way, but fundamentally the idea makes a ton of sense.
(40:00):
If I have, why can't connectivity just be more accessible?
why isn't the goal to, why are we marketing towards these couple of thousands of activeusers that are there in crypto?
Why are we not trying to reach to the Web2 audience?
People who are not in crypto right now.
(40:22):
I think we short sell ourselves when we say thatthe only way you can make a non crypto person come into the crypto industry is by way of
speculation.
Yeah, we did it in 2021, didn't work, right?
We did it with NFTs, didn't quite work.
The thing that I think should be observable is that when the president launched the memecoin, the market was able to absorb all of that liquidity coming through, which is
(40:53):
spectacular, right?
But so you have the attention, then it just went away.
Zero retention, right?
No way to retain that amount of influx.
And that's because I think a lot of the ideas are just still very crypto focused.
It's a lot of people within crypto trying to sell to other people that are there incrypto.
(41:17):
The total net new addressable user market never increases.
It's just the same people who are inside.
Then yeah, every cycle there would be some new speculative bet that could come up, whichwould attract some more people.
But you also have the outflow of some existing people who are not there anymore.
(41:38):
So we are net-net.
I think we are not adding as many more users as we would want to.
Now I think, is that paradigm going to change?
I think yes.
I think there are lot many good things that are happening.
happening.
But in terms of just the product, the low-filling product, I would love to have productsthat won't look good to begin with.
(42:01):
This is something I've heard about our product.
from a lot of people that even community members, this doesn't look like it's a Web3product.
And like, why should it?
If ultimately...
it's giving you the outcome that you want.
It's giving you an agent that is autonomously playing well, and also you are engaging, orit's giving you a hedge fund where you can see what the agents are doing without
(42:29):
actually...
any of the transactions or dealing with any of the wallet and all that operation yourself,why wouldn't you use it?
And that's why, that is one thing behind why we chose Hyperliquid, because we had some ofthese ideas back when we launched even the first product.
The vaults are a great example.
I can just directly participate in a vault without having to worry about anything else.
(42:51):
So that's like the design choice we made for Spectral.
Like make it super easy for anybody to use.
Just come and deposit into the vault externally.
So no, to worry about.
Am I giving my agent the access to my wallet?
Is this safe?
Blah, blah, blah.
None of that stuff.
Obviously, I mean safe here in terms of...
(43:12):
wallets and security, I don't mean in terms of profit because that's like the agents aretrading autonomously.
There's no way to say anything about that.
so yeah, I think that thinking needs to happen more.
(43:33):
And crypto needs to go, the crypto needs to become a boring back end tech if you reallywant to.
I agree with you and I'm really impressed by You know, I've had this many I've had manydiscussions exactly on this topic of how to make things just easier for just normal
(44:01):
everyday peopleAnd you mentioned wallets and I met with two different wallets, DeFi.app, as well as
Infinex.
And they're abstracting many of the things that we've kind of taken for granted as a pain,that having to bridge an asset from one chain to another, and then being able to transact
(44:24):
with that asset and then bridging it back.
Just that workflow and that operation is so painful.
What I'm seeing, a pattern I'm seeing is that many of these wallets are now, they'reencapsulating all the complexity behind all of that workflow and they're verticalizing it
inside of the wallet.
so, which makes it easier for the user.
(44:47):
Like I don't need to know, you know, about the bridging process.
I don't care.
I just want the bridging to happen.
I just want to switch from one asset to another.
Just get it done.
And I like that type of thinking because it'sWe're moving away from complexity.
It's still there.
It's just hidden.
But it removes a cognitive load that we have to deal with.
(45:10):
it's a lot of load having to deal with, know, bridging should never be so full of anxietyand fear.
You know, it should just be, just get it done.
Yeah.
And you know, even but the thing is like anybody who makes that's why like building incrypto is also hard because to make the right choice is not going to come with incentives,
(45:32):
at least not immediate ones.
So that I think that I think is just some bold decision making that needs to, I think,happen more within the space.
Like, know, if you go back to Apple,15, 20 years back, saying that no keyboard shouldn't be there on a phone, we are going to
think differently was bold, really bold, right?
(46:00):
With almost no predictability around whether this is going to work or not, right?
So I don't think like, that's also one thing that like I've just seen, like in the cryptospace, we need more of that bold decision making as well in terms of how we make product
choices.
It's usually the small things.
(46:20):
It's usually does this button have to be there or can you think of it differently?
Is there a scenario where this button is not at all needed?
Let's just get that out.
That I think needs to happen more.
So, yeah.
I love that you have your product hat on because we have a lot of devs in crypto, but nota lot of product people.
(46:43):
And I like that you come with a product background because you're thinking differently andyou're thinking more from the user experience perspective, which we need so much more of.
Yeah.
Let's go back to Spectral and tell us about 2025.
What are some things that the audience can look forward to and how they can get involved?
(47:06):
Yeah, sure.
So I think going forward, what you're going to see is just a lot of multi-agent tech, likemulti-agent native use cases come to Syntax.
Because every company, like if you think about it, Syntax is a marketplace of thesemulti-agents and agents collaborating with themselves.
(47:27):
But essentially then what happens is every company is a product by itself.
It's actually doing something for you that you can do on a, use on a regular basis.
and I feed something out of it.
Now,The Polymarket Hedge Fund, it's a poly and the fund is called Omen, and Spectre Hedge
Fund.
All these are just like, we went ahead with the use cases for specifically centered aroundDeFi and trading because that's the number one use case that you would want to put forward
(47:57):
and everybody uses that, right?
But going forward, this will start getting applied to a lot more nuanced use cases likecontent creation, creating content for your own AI models.
Like right now,If you want to participate in the creator economy, you have ideas, but you necessarily may
not have the personal persona of doing something like that.
(48:20):
Why can't you create, you know, why can't you create AI influences right now where you mayhave ideas, but you don't have to necessarily do all the things yourself.
This is by the way, I won't say more, this is actually something someone is building ontop of Spectral right now that goes on these lines.
(48:41):
there will be just like, you know, a lot of you faces coming around us.
And then going forward, I think where we want to take this is what we are going to achievein spectral as a platform out of this is we are going to be the network that has some of
these more capable agents, that actually do something.
(49:01):
Then our next challenge is how do we, so we have this agent name service right now, whichallows agents to connect and collaborate with each other.
The next part is how do you make this even more seamless for a retail user?
Like, can a retail user basically come and call 15 agents if they have a task and agentscompete with each other to complete that task and tell the user when it's Then you can
(49:24):
just tell that, get this done for me.
You see a plan created, agents compete, get that plan through, and you're there.
Or your flavors of product like that.
That's the initial experimentation going on and some builds going on internally and we areseeing what makes more technical sense and is technically feasible.
(49:47):
But that is something people will see coming afterwards.
That sounds exciting.
You know, I'm reminded by my conversation with with fleek because they asked the samequestion and they said that some of the most interesting
use cases they're seeing are actually from content creation agencies.
Which was surprise, which is totally surprising to me.
(50:08):
I expect the trading and the hedge fund use case because that seems very, just veryobvious to me.
But the content creation one was not so obvious.
And I think that's pretty fascinating.
Yeah, I think the real unlock comes becauseThe human brain is really good at thinking multiple ideas in a particular instance, but
(50:31):
not really good at diving down deep at the same time.
we don't have like the human brain just generally doesn't have good parallel processingcapability unless you're Sherlock Holmes and you have this, you know, mind palace kind of
thing.
(50:53):
Where agents actually come to great fruition is if you are still the idea driver.
I think nobody can take that away from humans.
are genuinely good at creating genuinely original ideas.
But then getting AI agents to parallel process those ideas and unleash creativity in thatparticular domain, that's where AI is most successful.
(51:18):
And you could see this in Ghibli, the whole Ghiblification trend that is going on rightnow.
Could an AI have thought of Ghibli?
as a style of animation, very hard, very hard to do, right?
There are almost no models there.
And I see this on a daily basis.
(51:38):
even if I screen record myself and computer use myself, we've done this internally, by theway.
I've done this where you screen record yourself over a couple of days and tell AI to giveyou new ideas of how you can be better at your work.
Doesn't work.
because I'm glad it doesn't because that means you would still have a role to do.
(52:02):
Right.
But that's hard.
But now you're seeing this with Ghibliification.
once, once you in space of art and space of content creation, once you have figured outwhat sticks, a particular style of content, a particular style of animation, it just
becomes a manual process after that.
There's no amount of creativity left in it.
(52:24):
So AI is getting really better atjust replicating those steps and giving you the output that you need.
I guess that's the reason why, and that's why this existing project that's going on istargeting content creation.
Let me hear Kulkarni from Spectral Labs.
Thank you so much for taking the time to spend with us.
(52:47):
Absolutely.
I'm glad we had, thank you for giving me my diatribe.
I'm glad we had like an open and honest chat.
awesome.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Betrayal Season 5
Saskia Inwood woke up one morning, knowing her life would never be the same. The night before, she learned the unimaginable – that the husband she knew in the light of day was a different person after dark. This season unpacks Saskia’s discovery of her husband’s secret life and her fight to bring him to justice. Along the way, we expose a crime that is just coming to light. This is also a story about the myth of the “perfect victim:” who gets believed, who gets doubted, and why. We follow Saskia as she works to reclaim her body, her voice, and her life. If you would like to reach out to the Betrayal Team, email us at betrayalpod@gmail.com. Follow us on Instagram @betrayalpod and @glasspodcasts. Please join our Substack for additional exclusive content, curated book recommendations, and community discussions. Sign up FREE by clicking this link Beyond Betrayal Substack. Join our community dedicated to truth, resilience, and healing. Your voice matters! Be a part of our Betrayal journey on Substack.