Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Token2049 Recap, Highlights, Insights
Recorded: May 8, 2025 Duration: 1:08:09Space Recording
Short Summary
The recent AMA highlighted significant advancements in the crypto and AI sectors, showcasing new project launches, strategic partnerships, and upcoming token generation events. Key players like GT Protocol and Fetch AI are leading the charge in integrating AI into crypto, promising to enhance user experiences and drive growth in the ecosystem.
Full Transcription
Thank you. mic check one two
hello hello guys. Hi.
Awesome, awesome.
I'm very, very excited to host this AMA.
So adding right now all the speakers to the space.
Give me just a couple of seconds and I'm welcoming everyone.
This is the first synth ama um centers decentralized
uh re-powering autonomous agents that solve uh real world problems and uh the main topic today
will be is a token 49 token 2049 insights was the updates was the um um insights and everything that you guys would like to share.
I will add in the speakers shortly,
and I have amazing co-host partner,
GT Protocol that will help me to run this AMA.
Hi everyone. My name is Anastasia.
I'm from GT Protocol.
I'm super excited to be here on X X-Space for our Token 2049 ReCup.
Big sense to you, yeah, for the warm invite. It's so nice to see here all of you guys familiar faces
and to meet some new ones too. Yeah, Token 2049 was an extensive and truly inspiring experience
for me and for GT Protocol especially. They came back energized, full of ideas
and grateful for the incredible people we connected with. I'm looking forward to sharing
some of our key takeaways and our global vision from GT Protocol and hearing from your perspective too.
productive too.
Awesome, awesome.
So I would start with a simple topic because if you're starting with GT protocol, I mean
we had super amazing time in talking.
There was many, many events.
We took several events together with GT protocol.
So Anastasia, would you like to share
something more about what Synth and GT Protocol are cooking, what's on the
roadmap and things like that? Yeah, sure. Thank you for the question. You're right.
Together with GT Protocol and Synth, we are about to launch a game-changing
solution for both crypto and non-crypto users,
something we have been building for a while and I can wait to share with the world in the coming weeks.
At GT Protocol, our core focus is on the AI execution layer, building AI agents that don't just talk but actually do the work and our main priority is an ai powered trading agent
hands free agents that executes your support and future trades managing your portfolio so you
launch it once and it just works in your background for example now by integrating
scenes uh 3d avatars and voice command interface built on fetch AI infrastructure, we are making
these agents even more intuitive and interactive and human-like. I believe that this is just like
a beginning of human-centric AI in Web3 that everyone is waiting for, but only just like
talking, but we are going to do it in common weeks yeah just wait for this awesome awesome thank you so much
I'm very excited about that because we've been friends and partners with the
GT protocol for quite some times and I'm very excited what we can do together I
mean the main the main guests that we have right now is ASI ecosystem. We have amazing, amazing speakers.
So we have today Fetch AI, the veterans of AI and Web3.
We have Singularity Finance.
We have Assister.
We have Rainmaker.
We have Kite AI.
We have Eliza OS, OpenX AI.
And then we have AI startups that are pretty new and fresh that also working on AI in Web3 directions.
It's Trendance, Glades AI, Strawberry AI.
We have Dissendow here, and we have Liquify and even Sophiaverse joining us.
So the next speaker I would like to welcome Fetch AI.
It was a big pleasure to be at the Fetch Hackathon
in Dubai with Sint.
We presented an MCP breach demo that we built together
for ASI and we've been a part of the Hackathon.
We are building and cooking amazing, amazing agents together with Fetch.
And I would like to give opportunity for Fetch to speak and introduce themselves.
And we'll ask a couple of questions.
Hey there, ladies and gentlemen.
Thank you very much for the very healthy, warm welcome.
My name is Nico, and I've been a community manager and a part-time dev row and
now pushing into the kind of marketing lead position on the Fetch.ai ASI funds and I'd love
to, it'll be a pleasure to shed light and clarity on kind of what the project is, where we started
from. Maybe just one question. We started back in early 2021, prior to that actually, all the way
back in 2019, but let's say early 2021
was the inception.
We worked for about two years on a framework
called the Autonomous Agent Framework,
which later became the Micro Agent Framework in late 2022.
So a lot of the agents that were developed
as part of the hackathon in Dubai
that Synth were able to present that, where
all effectively new agents, so-called microagents, a lightweight Python framework that helps developers
stream agent creation process.
If there are any Python developers listening to us live, all you have to do is open up
your IDE, type pip install uagents, and you'll be able to think with the agents immediately and with that being said i'd welcome any questions you have both from sent
the host as well as from the community yes yes yes and i'd love to also hear a recap yes uh so i
mean yeah go ahead the first the first questions we have for sure i mean i'm very um interested in um si1 mini
and uh i mean we've been playing with it for uh quite some time already with same team and if you
can please share um a bit more information about si mini llm because it's pretty cool i feel like
it's the first uh web3 decentralized uh llm that kind of like go uh in uh competition with the main leaders on the
market like open ai um gemini uh and tropic so what's the key difference and um um you know
what's what's the key uh purpose and what's the key differentiate differentiation all righty then so ASI mini is a large language model we
created and announced uh back in the 25th of February I believe so it's a relatively new
product about 70 days old and it is a relatively small large language model about 39 billion
parameters optimized for reasoning and acting as an orchestration layer and what is it orchestrating
between other models so as part of the asi trade initiative which is an initiative
under the asi umbrella we're aiming to uh both integrate open source models and subsequently
fine-tune them as well as kind of train new domain specific models to be deployed across
medical domains and to be deployed across the
automotive and broadly speaking robotics sector and so there's a there's a desire for us to create
a so-called mixture of expert system which is the first differentiator asi mini is an orchestrator
model a relatively small model that needs to orchestrate between other models there are
experts specialists in their domains and the outcome or at least the desired
outcome we're hoping to achieve is increasingly high quality inferences and insights one ai model
orchestrating between kind of a network of other models to ensure that can to ensure that broadly
speaking answers the answers that the large language model generates are relevant and they're
grounded in facts and
they're as non-hallucinatory as possible and that they lend themselves to egetic execution
and so the second differentiator is that the large language model of ASI Mini is built for
really with agents in mind and the goal there isn't to have natural language in natural language out
which is what we see from most large language models now the goal is to have natural language in natural language out which is what we see from most large language
models now the goal is to have language in useful economic output out and so the model orchestrates
agentic activity in the back end and what we end up having is a very simple interface which allows
people to speak any intent they want into existence just type what you want maybe you want to book a
flight or maybe you want to book a flight or maybe you want
to book an airbnb maybe you want to buy a piano maybe you are interested in having a coding
assistant maybe you want web server try it you know ai is general purpose technology so the use case
is you know they're practically endless really the sky's the limit in terms of use cases and what we
have is a lot of agents sitting in the background with ASI Mini orchestrating between these agents.
And you have increasingly high quality inference as a consequence of the mixture of experts architecture.
And you have increasingly high ability to display kind of embodied intelligence, leveraging the agents as connectors to tools, your calendars, your design tools, your productivity tools, your coding tools, and so forth, and leveraging the agents as representatives of people, places,
and products, which effectively kickstarts the AI economy, right? You have a very simple funnel.
Users interact with ASI Mini. ASI Mini orchestrates agents in the background. The agents in the
background are built with the U-Agents framework, the one that we launched back in late 2022,
built with the uagents framework the one that kind of we launched back in late 2022
and they are living across the agentverse which is uh you can think of it as a port terminal
for agents a giant hub which helps developers with their hosting needs gives them analytics
gives them kind of soft debugging tools and the outcome there is to ensure that agents have a home
and that agents are discoverable and that agents are ranked and that
their behaviors are logged and that they are indexed right giving people the ability to
you know seamlessly interact with a critical enough mass of agents all of which are becoming
increasingly useful and this entire utility this corpus swarm of agents is neatly harnessed by a unified interface.
And so that is kind of the attempt there, technological, I suppose, to compete with the giants of the world.
Mixture of experts is, I think, a creative twist on an age-old scaling problem in AI.
And, you know, if you have infinite capital, then you can afford to create giant everything models that master everything from programming to pottery to poetry and so on.
But if you don't have infinite energy and infinite capital, then you're going to have to get creative with how you scale.
And so Mixture of Experts is a kind of variant scaling approach where one orchestrator model orchestrates between domain-specific models, therefore, you know, giving higher quality
insights. And yeah, I think that's the overall pipeline there. You've got agents as a middle
layer. You've got the ASI LLM acting as kind of a central interface. You know, if we've learned
anything from chat GPT's release so far, it's that humans really love the conversational interface.
For some reason, interfacing with your AIs
by simply speaking to them is just very intuitive for us.
And so that's also a secondary sort of benefit,
but still a higher order benefit that we get
from kind of having the large language model
because we no longer need to ensure
that people know how to interact with thousands of agents.
Now the entire thousands of agents orchestration process
is being handled by the model
and there's a certain grace and elegance there.
So this is my two-minute elevator pitch.
Awesome, thank you so much.
I mean, it was very, very, very informative.
In a simple words, you know, working with Agentverse,
working with Fetch, working with SI1,
I can definitely tell that, you know,
there's cool unique difference that you can actually tell that, you know, there's cool,
unique difference that you can actually not just send the prompt to LLM, but you can also send the command directly to the agent. You can call the agent directly from LLM,
and it gives the very, very big range of, you know, functionalities and possibilities what
LLM can offer for you. So I'm fully in into decentralized decentralized llm because i think it's the future you know
it's only one way to build a proper open source stack and you know the proper ai development as
a future we are with sind building a si bridge where you can call synth agent directly from
uh si1 llm and we since we will breach a huge ton of MCPs directly to the agent person to ASI Mini.
Thank you so much, Fesh. I think it was very informative. I mean, please stay with us and
you know, if we will have more questions, we will ask and send it to you. I really appreciate
it. So the next speaker will be Singularity Finance.
Hi, everyone.
A big shout out to Nico, by the way, because the workshop that you gave us,
I started making notes.
So loved it.
So my name is Sandra, and I'm taking care of partnerships at Singularity Finance.
And as the name suggests, we're a child company or like a you know ecosystem project from singularity net and i'd like
to always say that we are this financial arm of singularity net because while they are doing
no ai development very technical stuff we are trying to focus on the infrastructure specifically
the financial one and singularity finance is actually a result of a merger between two you
know companies cogito and singularity dow because at the end of the day we
were talking the same language of on-chain finance in different ways so right now we're building that
kind of on-chain financial infrastructure for ai companies and ai agents both with our tokenization
framework and with our vault frameworks which is more focused on on ai agents and i would say that after the token 2049 uh this
this area of ai agents was very you know bullish and helped us validate some ideas um that we were
working on so i'm very excited about this element specifically because we've we've been talking with
multiple of ai agents that are doing a number of things. And there's always this, you know, additional skill set
that they would be lacking, right?
To be really useful, really like boost their utility.
And yeah, I would say that this is like the rather short pitch
going to take Spotlight from Petch.
I absolutely loved it.
It's a pleasure to have singularity here on the
call and i'm very amazed you know having uh such speakers and partners together in one place it's
very unique ama thank you so much singularity a quick question uh so um uh as we as we uh speaking
right now uh you know we're also building with synth uhint our 3D avatar that can pretty much can simplify interaction,
simplify UX with voice and then give a face and emotions to any agent.
And that's the direction that we're currently exploring with Singularity Finances as well.
Anything you would like to mention, you guys building as a hot thing,
You would like to mention you guys building as a hot thing, kind of like the next cool DeFi trading or algorithm, something that you would like to share with us?
I would love to, but our team is actually preparing a roadmap, so I don't want to spoil the fun.
But you mentioned some things already I think that as I mentioned this this past week really
helped us understand the market and kind of adjust our strategies and where we should be going so
that you know we capture the market so don't want to spoil the fun but yes so I think that the
agent space is definitely place to be especially with that kind of comprehensive vault frameworks that we already
see a very big interest in. Awesome. Awesome. I appreciate it. Thank you so much.
So the next speaker, I'm very excited. It's ElizaOS, very hyped project, still on the hype, and I'm very happy to have the main lead developer from Eliza here.
Let me know if you guys are here, if you can speak.
For some reason I see that...
I think they requested some voice, but you didn't approve them. Я думаю, что они попросили голос, но вы не обеспечивали их.
Нет, я думаю, что это не конкретно.
Но мы ждем за Элайза, который еще здесь на AMA.
Также мы будем на следующий спикер.
Ник из «Ассистер».
Плесня, что ты здесь, брат.
Поговорить было очень весело, и мы были вместе на несколько событий. uh nick from a sister uh pleasure to see you here brother um like talking was super fun and we've
been together on a few events uh i know you've been with eliza as well on the same event too
um so uh nick what's the updates from you tell us a little bit about the sister and uh kind of like
some vision uh behind the agent incubation and i know you're building a lot, so please share any updates.
Yeah, sure, man.
It was a pleasure to meet you in Dubai
and a pleasure to join such a great community call,
let's say, right now.
It's great to see that we have around 7,000 people in.
So a little bit about a sister and myself.
We started building it at a previous bear market, which was tough, really tough.
We're a Cambridge-based team.
We used to work on AI stuff, ML stuff quite a long time.
And basically what we're offering, we have a robust ecosystem where we have AI builder,
where anyone can build the AI agent without tech skills.
Then we have quite a big marketplace with 4.2 million users right now, with more than 24,000
agents already created by our community and registered users. And this ecosystem is thriving.
So it's like we have a lot of usage. We believe that agents should be not only entertainment or only social,
you know, like which doesn't have any real value or like utility.
In the opposite, we believe that utility agents,
it's a new generation of agents that are going to be replacing SaaS businesses,
as we know, because the best thing that crypto can do
is actually tokenize some assets.
Because we have speculation, we have DeFi sector, and we have tokenization.
So with tokenization of AI agents, I believe, of utility agents,
I think that's where the huge opportunity lay down.
And that's where a lot of people can build working business, investment portfolio,
also participate in different incentives. build a working business, investment portfolio,
also participate in different incentives. And we want to be a platform that helps to find right agent,
coordinate agent to agent in common work.
And we want to be a platform that helps to discover
this investment opportunity for people who want to get exposure
into a genetic investment
portfolio. We expect to have some TGE, I believe, in next four or six weeks, hopefully. The market
is getting better. We did two big events at Token 2049, so that's where we met with the Lysa guys,
with a couple of other guys. So we had DEI Agenda Day, where we met with elisa guys with couple other guys so we had dei uh agent day
where we host a lot of deep in projects which we collaborate with which we partner with which we
have very strong um collaboration around data um and on other side we also had uh decentralized
ai agent day where we had a lot a lot of partners invited and chatted and shared a lot of
insights about the future of agents, how they're going to be cross-collaborating, how they're going
to be sharing profits, how they're going to be discovered, how to choose the scoring of agents
and their ideas are verifiable and on-chain.
I think a lot of this thing is something that we as a team care a lot since 2022, 2023.
And finally, we see a lot of partners and teams building in space,
delivering agents, delivering different protocols,
different frameworks that are allowing us to organize that in really efficient way, first of all, and second, with good UX UI experience, and third, it's finally happening. So thanks a lot for
inviting again. We'll pass back. Thank you, Nick. I mean, a sister was crashing already for quite
some time on the AI market, and I'm very happy for you and very happy for success and the launches you guys did.
So amazing having you here and hear updates and, again, discuss the collaboration
because I feel at this AMA, it's not just, you know, it's not just to share the news and share the updates,
but we all can build something together and create some unique use cases that
users gonna like and enjoy and actually have some profit or
utilization as you said so um i'll give the mic to the next speaker the
next speaker is the kite ai kite ai is a well famous project on the ai space
and we have a bill bill as the speaker at this event.
So please say some cheers.
Thanks so much, Sintim, for inviting us.
So my name is Vik and I'm the head of protocol at KiteAI.
So at KiteAI, we are building what we call
as a AI-first purpose-built L1 chain.
So we are working with the Avalanche ecosystem.
And we are building this compatible L1 blockchain,
which is a purpose-built, like I said, for AI agents.
And the goal is to create this marketplace where people could deploy agents, models, data.
And then based on these data model and agents and other AI assets, we also want to provide composability where builders or partners or enterprises could assign identity to their
agents and give wallets and then we are have also worked on a proof of concept with
smart contracts and slas between agents and we're working with some of the known names in our
industry like eigenlayer walrus and others for verifiable attribution
for inferences and AI agents, and also working on some TEEs for privacy and security-based execution.
Awesome, awesome. I mean, it's very interesting. I mean, just like one question I would like to ask is like, what the EVM current features you have with your agent?
What are you covering? What's the main use case and functionalities?
So we are not building agent ourselves, but we are bringing all these partners who are deploying agents with us. So we have at least three agents on the,
so we are in V1 testnet,
we'll be launching our V2 testnet in a week.
So there are quite a few features
that these agents currently have.
Three of our partners, Bita AI, Bitmind,
and Codata are the current agents.
And we have also deployed our own agents for testing purpose,
dog fooding, if you will.
So these agents have multiple roles.
So if I can tell you what these agents would do.
So every time an agent is deployed, like I said,
the agent gets an identity.
So this identity also has a verification layer. So we want to verify if the agent gets an identity. So this identity also has a verification layer.
So we want to verify if the agent is not malicious,
whether the agent is coming from the right owner.
So we are also looking at how do we provide KYC and KYB
to the owner of these agents so that the agents
and the task that the agent can perform is trustable.
So we are trying to provide trust as well
with some of our partners in that area.
And then like one of our core uniqueness
is proof of attribution.
So every time agents models data
or any AI assets on the platform gets utilized
to build something or to for the
consumer when the consumer consumes any services. All of these information is recorded on chain
and based on the price of each of these services, very similar to the chat GPT approach, there
are token price per token for all these AI assets. We capture all the information,
which is what we call as attribution on-chain,
which is verifiable by anybody because it's open source.
Based on that, we give rewards to the contributor of these agents,
models, and data and other AI assets on the platform.
Awesome. That's super cool.
That's actually the. I mean, like that's actually like the functionality
that definitely needed on the agents market.
Thank you so much for sharing this information.
And I would like to invite the next speaker
and make some rain.
I know Rainmaker for quite some time.
The guy's been on the market for several years at least.
Definitely I know them.
So we have JD, founder of Rainmaker.
So JD, please introduce yourself
and maybe share some more information
about what you guys do at Rainmaker
and what's the beauty behind it.
Yeah, thanks for having me, Ilya.
And I guess in crypto years, I'm basically like 100 years old now.
So, you know, I started this journey about seven years ago with a passion for how can
we empower humanity with the combination of AI and web-free technology.
And that's really been the core focus and still is the core focus today at Rainmaker.
When I look at the future of where we're heading as a species
and the role that AI has to play in the evolution of humanity
and the evolution of human consciousness,
there's a tremendous responsibility, not only for us as a company
and for me, but everybody who's building in this space.
We have a responsibility to lead
humanity in the direction, hopefully, where we can fully empower them.
So in terms of the company, we focus very specifically on the training of models and
agents.
So data validation that's powered by a system that we've built called Train with two eyes,
just like Rainmaker with two eyes.
So any model or agent
can essentially receive data validation or support in their training through the rainmaker network
and we've got almost a half million contributors and validators on the network so right now on the
agent side we're actually we're gonna i'm gonna drop an announcement here that we haven't even
put out yet we have a product called train agent that we announcement here that we haven't even put out yet. We have a product called TrainAgent that we've launched, that we're launching rather.
And the first application of it is actually a native plugin that we built on Eliza OS.
Since I figured since Eliza is here, we might as well drop it, put some alpha in the chat.
And so that plugin is going live, I believe next week on the ELISA framework. So any agentic system that wants to access data validation
to help support the training of their agents by Rainmaker,
they can do so natively through the ELISA framework,
the plugin that we've developed in partnership with ELISA.
So that's coming next week.
We also are doing stuff, like I said,
around data sets to train any model.
We've got some other cool product stuff
rolling out this month.
And then we're actually also heading right into our TGE,
which is just like six years in the making.
So like I said, I'm pretty much a masochist to be still at it for this long.
But thanks for having me, Ilya.
And I'm excited to be a part of this group today.
I mean, if you look around the room,
you've really put together an esteemed group of companies that are all at the forefront of, of launching centralized AI.
When I, when I look at the landscape, like people often say, well,
why are you a competitor with this project? And I'm like, look,
in a decentralized space, we are all, none of us are really that massive yet.
Hopefully we'll get there. I mean, our,
our competition is centralized systems that seek to enslave humanity and to further solidify the power structure that's already taken us to the brink of where we are as a species.
So I look at everybody on this call, whether we're doing similar things or different things, if there's a way I could support any of these companies or other companies that are listening, Rainmaker is here to be of service to the centralization movement.
So thanks for having me. Awesome, JD. Thank you so much. It's a pleasure having you here and a pleasure to have all speakers because as you say, it is definitely companies that building,
creating not just the hype, but actually real solutions on the market. The next project is
quite interesting, but before that, I want to share the um that eliza
shared with us uh sorry because they have some network issues right now to connect into ma but
we we will have them on our next sessions they will definitely uh be here but probably for the
next time there is you know internet or connection issues it's always happening uh last minute i would
like to welcome the next speaker is Strawberry AI.
So Strawberry AI, really cool name.
Give us some some pitch and information
what problems you're solving and what you guys doing.
Yeah, hello.
Nice to have you, everyone on board.
And yeah, we covered you on Strawberry AI.
We started like nine months ago with a simple idea.
And that was simplifying DeFi by leveraging AI.
It was like nine months of hard work because we made everything from scratch.
And for that, I don't just mean agents.
I mean, actually, LLMs and hosting infrastructure up to the inference engine
we have a strict policy of doing everything in-house because we want to optimize everything
especially for like d5 evolves that that are not easy like if you try every centralized ai and having it deal with ai like with defy
it's gonna be tough so we spent quite a few time like uh running our own pipelines for data scraping
and i mean off chain and on chain data scraping ranging from like twitter uh up to like token holders prices and now it's everything in place but the end goal
was not just consuming data and serving it for agents it was just for one simple reason and we
wanted to build from the very first beginning the first AI agent that lets you easily trade
the first AI agent that lets you easily trade any crypto you want,
no matter the chain, no matter where your funds at, by just prompting.
And I'm proud to say that we reached that goal,
even though it's still far from the overall picture that we want to build.
But yeah, during during token 2049 we
launched very swap which is our our own like flagship agent that lets you basically i i call
it like by vaping because it lets you trade any crypto without you don't have to research so much
i mean the heavy lifting lifting is done by the llm itself and the
agents including like researching if it's a scam or potential red flags the holders and up to the
prices and it's a good entry or a bad entry all of that basically all the the life cycle of a dgen trader it's embedded into a defy agent and it lets you
trade and having your sort of ai wallet that you can use in our everyday life actually
it's already live in prod and people are using that but uh we want to make it like a step further
that but we want to make it like a step further right now we're working on like proper chain
abstraction even now with for example with spectra we're we're thinking about like uh avoiding most
of the like uh let's say painful steps of like end users like having to approve stuff we want to just have for the end user is simple to use
prompting interface and eventually also voice interface to just buy sell set limit orders
ape into stuff without even caring much that's our end goal and eventually the long-term goal is having an AI-native wallet that runs into your own platform that lets you do crypto life every day with an AI assistant.
In this case, it's an agent that maybe will recommend suggestions.
It will let you trade.
And yeah, that's what we're focusing on awesome i would say we're
focusing more on users rather than infrastructure even though we are infrastructure heavy yeah users
always are the key so um pleasure to having you here uh again as i said you know the this space
uh is not just about to uh share about what we, but also find the ways that we can
bridge and build things together.
So I'm looking forward to reconnect
with everyone after the AMA.
The next guest is a really good friend of mine,
Ashton from OpenX AI.
We had amazing meeting on SI event in Dubai,
where Ashton was presenting super cool tech uh what they built uh uh with
openx ai um and share this like a live demo of speaking agent that was super cool ashton i mean
everyone loved it i would like to give you a word and share a little bit about openx ai what you
guys building was the core difference and how you actually build that amazing tech together
was the core difference and how you actually build that amazing tech together
cool cool thank you thank you for having me um after so many meetings uh meeting with people
and talk of 49 i thought i'll be done with meetings but here we are uh it's still going
very strong so excited to see i think how many members we've got uh almost um 7k plus
which is amazing to see so so thank you very much for having me um great to see many people
are working towards to uh in my opinion basically fighting the centralized vertical integrated
monopoly in ai and infrastructure so it's great to have everyone.
So OpenXAI, it's a permissionless AI protocol. We've been building for quite some time now.
So I wanted to maybe give you a differentiation
how we see a permissionless AI, how we define,
because everyone is familiar with a closed source AI tech
like ChatGP, ChatGPT, and open source AI,
maybe services like LLM models.
And then you have a decentralized AI
where you have lots of computers and people,
they basically run things together.
You have a bit of a governance,
you have a federated
governance or or maybe centralized governance so the idea here you have no central single company
or controlling it so permissionless ai to me is basically basically a methodology or a concept
that anyone can use it anyone can build it similar like a bitcoin network where you just need your
computer and internet connection and you can host your ai services and rent your ai services so um
a lot of these ai modules can be deployed onto decentralized infrastructure rather than having
like aws because i see still like in the decentralized space,
if you look at a lot of services,
you still have to register using your traditional methods
like email address and at least signing up
with their process.
So permissionless AI to me more like a way you can access
without going through any KYC process.
So, yeah, so we've been building these protocols for a while.
We had an event in Dubai.
That was basically a fun event for us because we were showcasing how a technology, not just, you know, sometimes technology can be boring.
But what we did, we had an AI kind of
agent panel, we worked with Fetch AI, new agent framework. So these agents, they were debating
each other with a specific trained personalities, we had Donald Trump voice, and we had Michael
Saylor, we had Peter Schiff and also Mysterious Satoshi.
So the funny part is these AI agents, they were placed on a chair in a vertical screen.
They were chatting each other. So it's nice to see because Token49, a lot of events, people were sitting and talking.
So in this event, AI agents debating each other in real time.
So it was a really good experience and also for us to showcase our technology.
That's awesome.
It was actually super cool.
I pinned the video from Openx ai account to this ama so
everyone can see that i'm super excited to uh integrate uh tech that we build with fetch and
open x ai together with sin to make it um not just uh speaking fun um but also um have our avatar
and emotion set and everything um i also, remembering your Mimic AI,
that's also super cool when it can actually identify
the accent of the person and speak in exactly the same language.
I had amazing fun experience, live experience,
communicating with the agent during that panel session.
So I really appreciate you, Ashton.
Thank you so much, OpenX AI.
Please check it out, the project.
And we're looking forward to have more cool things
what we've been sharing and building together.
I would like to invite the next guest.
It's Trendance AI.
Super cool, new, interesting project
that I'll be happy to share and give opportunity to speak. I think it's very interesting. It's an
orchestration layer. So Trendants AI, please share.
Hi, Elia. Hi, bro. Thank you very much for the invitation to
this amazing Twitter space. So my name is Max and I'm the
founder of Trendens AI.
Trendens AI is an AI orchestration layer.
So basically we integrate multiple LLMs, AI agents and APIs into one unified infrastructure
and building a mastermind AI on top of that, that will orchestrate and execute users prompt in the most efficient and accurate way.
So very interesting prehistory to this that I have been building
and incubating multiple projects for the past three years under the zero stage.
And around half a year ago, after all this AI agent boom,
after all this AI agent boom,
the good idea came to me.
And I see that a lot of agents
is becoming more and more in our space.
And more than 35,000 agents have been released.
And I see that it's very hard for people
to find the best way,
the best agent to execute their
prompt like numerous of agents on fetch ai as they released just recently like more than two million
and i think that orchestration of agents um building like ai on top of other ai agents is um
on top of other AI agents is very, very like emerging technology right now, because users
don't need to find like numerous agents to execute. They just need to talk to a mastermind AI,
and the mastermind AI will choose the AI agents for them agents for them. So it's very interesting, maybe it's a coincidence,
that Fetch AI also is doing quite similar thing right now.
And I think it's a very, very good sign
that we're moving in the right direction,
that big players like Fetch, that we, by the way,
already partnered with, are in in one direction and right now um we are looking
at like researching like uh the best use cases how to unite like crypto world and real world
and ai agents i think can really solve and tap into this kind of audience
and then board many people to the crypto
and also make some kind of a bridge like into crypto and out of the crypto
because it's still very immature and it's still very complex for many people who just hold like i i
call them usdt users uh they don't trade they don't uh sell they just receive for example salary
in usdt and this is very very huge amount of audience and we're right now exploring like
huge amount of audience and we're right now exploring like use cases that can target this
specific user base and yeah just help them to use crypto and real world awesome awesome thank
you so much max thank you so much trenders please check it out very cool project um uh we will be
uh signing the partnership uh pretty soon and uh posting into
in into the social media about it uh looking forward to uh build cool things together uh the
next project is not exactly ai but something that can benefit ai ecosystem it's a web3 project um
liquefy liquefy please um uh do an introduction.
Hey there, can you hear me?
Yes, yes, yes, we can hear you.
Hi, Andy here from Liquify.
Thanks for having us and shout out to all the speakers around.
Love to see so many great projects and people around.
As a short introduction about what we are doing, we are making illiquid assets liquid on a short notice.
So basically the tokenization of capital that sits idle. As you mentioned, we also were present
at Token2049 and it was a blast, but in the business development side of things, like meetings
with partners and clients and whatnot, it was a blast. On the other side of the story,
we are not quite AI, as you mentioned,
but we are from the DeFi.
And our mission is to kind of go to the DeFi 2.0 sector,
making the liquid assets liquid
and maybe solving some of the liquidity issues
that we previously had and are having in the web sheet.
Also on the other part of the AI, shout out to Fetch.ai here,
we also are integrating one of their agents on our platform.
So whenever we are making those illiquid assets liquid
through a tokenization of soft agreements or staking rewards,
APRs and what other idle capital you're seeing,
you will be able to also make more informed decisions
to those agents.
So yeah, this was a short introduction about us.
I hope to see many more search spaces
because you guys did a very good job
on the organization part of things.
Really, really good shout out for you there.
And thanks for having us. Thank you. job on the organization part of things really really shout good shout out for you there and
thanks for having us thank you thank you so much as uh liquefy is a partner and partner in esi ecosystem so we're definitely um very thankful for you to be here again it's not just about ai
right now but it's about how to breach all the tech together and how to build the best tech
for uh for the customers
where we can solve real problems.
We don't have much time,
so I'm trying to squeeze it as much as I can.
So the next speaker is Glades AI.
Please explain a little bit
what you guys do on Play to Train
and kind of like what the top features you guys do
and what's the difference.
Yes, yes, thank you for having us.
So hello, everybody. My name is Robert. I'm the lead
developer and founder of Glades AI. So what is Glades AI? Glades AI, we gamify AI model
training. To put it very short, very simple, we gamify AI model training by generating
video game assets, such as map terrains or bots in the game.
So essentially, companies give us big data.
We turn that data and our neural network architectures into these different game assets.
And then game developers implement our platform so that gamers can use our game assets in the game.
And so as they interact with the game assets in the game, the gamers are actually
interacting with the AI, giving feedback to the AI, training that AI. And so we utilize these
gamers' GPUs to train this AI, and we give them a piece of the pie. We give them a financial
incentive, a financial reward for training this AI. And because we're able to do it at a lower cost than these
big companies, we actually give these big companies a discount for training their models. And game
developers have the extra incentive to use us because gamers, now they will also adopt these
games because they can earn a financial reward instead of just having fun. In addition, the game
developers can save time and money
because the assets that we generate for them,
they don't have to generate themselves
or pay someone to generate for them.
And so we want to create this ecosystem
of a win-win-win relationship between gamers,
game developers, and these big companies.
Basically saving everyone some time and some costs, adding more variety
and diversity into gameplay, and allowing gamers to actually pay off some of their games,
pay off their GPUs, their very expensive hardware, and paying off their electricity.
And so that's Glade's AI, gamifying AI model training.
That's super cool.
I'm very amazed with you guys building.
And to be honest, let's discuss the next step with Synth
because Synth is not just an agent.
It's also an AI NPC, as you understand.
It's a 3D character that could be integrated into any games.
And looking forward to discuss what the bridges we can do guys together.
We don't have much time left.
I would like to also run some Q&A sessions from the crowd,
but I want to give the word to our partner and really cool project that also AI and also gaming.
It's Sophiaverse. Sophiaverse was on the market for quite some time and building really, really
cool, amazing metaverse technologies. So I would like Sophia to give the word what they currently focusing on
hello everyone you know what they say the best is saved for last uh thanks to ilia uh sent the gt protocol for organizing the space and the invite shout outs also to our ecosystem partners
fetch ai and singularity finance uh hey to all other participating projects. It's nice to meet you and to everyone listening.
Thanks for being here. I'm here to tell you a little bit about Sophiaverse, our core product,
the sentience AI lab and some of the AI tech we are building. The sentience AI lab or SAIL for
short is essentially a gamified AI environment where players interact with AI in fun, engaging
and rewarding ways. It brings together a mix of advanced AI systems and immersive gameplay.
Players interact with AI NPCs, each with their own personality, memories, emotional states, etc. And as players interact
with them, their behavior evolves based on the player's action and dialogue, making
every encounter feel personal and unique. To achieve this, we are building several AI
systems. First, I want to tell you about the AI cube. It's our backend system that contains many of the AI
systems and it acts as a bridge and manages how AI handles
conversations, memory responses, etc. in the game environment.
It's assigned with a modular structure which allows us to
easily plug in new AI
technology into what we call the orchestra.
Then we have two key systems.
One is called the sentiment, which lets players have freeform unscripted
conversations with AI NPCs while also giving the player clear choice points
and objective-based progression.
And it plans talking with a chatbot with classic forms of conversational gameplay
used in RPGs to advance the story and develop characters.
The other one is called Deed that analyzes player actions or deeds, how we call it, to influence how AI NPCs react.
You could even say how they react emotionally.
If you're nice to the AI NPC and you treat them kindly, they will want to be your friend.
If you're mean to them, they might give you a harder time at progressing in the storyline.
you're mean to them, they might give you harder time at progressing in the storyline.
On top of that, we are using Singularity Nets Iris technology as foundation for our experiential
learning system, which gives AI NPCs the ability to follow their curiosity, explore the environment
with full agency and study the player's action to get better at completing
tasks over time. And we just announced the phased rollout of the sale alpha starting
this month with parts of our community. And we would love to invite all the projects here
to get some alpha spots for their communities or for yourself so if you're interested feel free to reach out and we
can arrange something that's a quick rundown of sophiaverse again great lineup the organizers did
a fantastic job and uh thanks for having us and looking forward to connect thank you thank you so
much to be honest i'm super excited i don't want to share too much, but I'm so pumped to see Synth in the side of the metaverse,
Synth AI NPC and then also utilizing
the Sophia with Synthag that we're currently building.
We've been partners since the beginning,
I would say, and definitely, you know,
in a good space to build some really
cool use cases for gaming.
So guys, the introductions and kind
of like brief updates are,
I'm about to end.
We have like around five minutes left.
We got almost 10K users.
I think it was 8.5K at the max.
So for like for the first AMA, I think it's super cool result.
But I'm really thankful for the partners and speakers who join here.
I would ask Anastasia from gt
protocol to help me with q a session you have very unique opportunity to ask fetch singularity
openx ai a sister glade sophia verse trendens you can ask questions you like uh so um please uh gem
yeah thank you yeah i already collected some interesting questions from our community
chats so the first question i think goes to fachy and singularity um the question from the bullish
pizza guy were there any under the radar projects or startups at token 2049 that caught your attention what made them stand out and
why do you think they are worth watching
uh singularity which one go ahead go first uh as you wish for the first one
uh as you wish for the first one
all righty then well I wasn't personally part of talking 49 so I'm a bit uh kind of skeptical in
giving context around it but what I was able to catch is uh open xai's open protocol and their
ability to seamlessly connect conversational agents I loved it I think uh you know people
might see it as a as a kind of toy project, but there was a deep insight,
and that is that interagent communication is no longer a bottleneck. And so if you were to take
that openness and kind of seamless agent-to-agent connectivity and connected with something like like Iris building on Sophia versus integration. Iris is interesting in that it's, it allows,
it stands for autonomous, intelligent reinforcement,
inferred symbolism.
And it's basically a self-play model
that understands its surrounding
by observing and analyzing changes, right?
And so you can already see how something that is able
to basically learn from cause and effect,
because that's what it is
it creates it derives rules casual rules from its environmental observations and it encodes
incremental state changes and these rules dynamically interconnect to form an action
state graph right which serves its its internal world model so we have agents with an internal
world model that are able to seamlessly connect with other agents and can already see how that kickstart the ai agent economy i think what token 49 at its core signified is that everyone
is building towards interconnectivity of agents uh in in in some form or in some capacity or another
and once we have interconnected agents it's only a matter of time that we have the agentic
reputation systems and we have the
seamless connectivity with tools through for example retrieval augmented generation
architectures and so we can already see how there's actual impact ready to be had countless
amounts of use cases and i'm really looking forward to every other token 49 that follows
because i can already see the explosion in kind of creative awesome awesome thank you so much yeah so i'm not gonna speak uh for fetch for you guys but you know i
can speak from the scene angle and i know that we will be at the event on 17 of may in um in silicon
valley where uh fetch will be uh presenting the uh mcps on AGI House and since currently building the MCP bridge that we
will be bringing bridging to Agentverse and ASI Mini that's kind of like the one of the coolest
updates so please check out for the our news our channels to see what's happening and whoever in
Silicon Valley who are in Bay Area or SF please join us to this event. I'm at it. Java is one of the speakers from the community
who wants to ask one question. Java, please
ask.
Thank you for bringing
me up. My question is
to Fetch AI.
I would like to know what
challenges participants faced
when they were building AI agents
during the Fetch AI,
Fetch Global Hackathon?
And how did they overcome such challenges? Thank you.
So, Gary, a very context-specific question. I think I wasn't directly leading the hackathon,
so the best answer is I have to get, I have to, I have to sync with kind of one of my, one of my colleagues to give you full context.
But what I've seen as multiple barriers is work with Python virtual environments.
People that aren't familiar with Python kind of struggle with that initial step of setting up the virtual environment once you're there.
What we effectively have with the UAgents framework is kind of a framework that allows generic asynchronous message passing.
So it's built on the UA agents in part are built on an async IO library, which is, it opens up the door for scale because, you know, the asynchronous message passing gives you concurrency and whatnot, but it's also a challenge and it a kind of a new way of thinking for a lot of devs so there's a that initial conceptual kind of hurdle that that from an
educational perspective uh from there it's not just you agents right we're kind of building an
integrated ecosystem so it's not that they only need to learn how to work with the framework it's
them needing to understand how how to work with agent verse them needing to understand how to
work with asi1 and kind of what are the integration peculiarities there so a combination of kind of conceptual
challenge and practical bottlenecks but usually hack not participants see it as see these as kind
of trivial tasks and all kind of very expected you know if you're going to onboard yourself to
a new framework and kind of new ecosystem of tooling there's probably going to be some friction
to start with usually within 30 to
40 minutes everything's been troubleshot and you're set to start then yeah that's awesome thank you so
much i mean hackathon was pretty cool to be honest and the organization of hackathon went super well
i think we as a scene participated and built a super cool mcp around fetch tech and we build a
meme generator that we will be launching super soon.
So please keep eye on the fetch and send socials.
I will give another question.
I know we have really, really kind of like a last minute,
but we have a professor who would like to ask
question from the community.
Professor Morty. I guess Professor Morty has just got disconnected. Any other questions we have? Anybody wants to ask? I found one question from the community, I think it's
a general question for all our super speakers. I'm going to ask
What do you think guys are the top areas where AI agents can truly
revolutionize the way users interact with crypto platforms?
Whether it's onboarding, trading, support or beyond, where do you see the biggest potential for AI to change the crypto game?
potential for AI to change the crypto game?
I have my own question but I will give the words to the people.
I think Nick wants to answer immediately.
Okay, yeah, let's go.
Yeah, I can just drop two things.
It's like today I was sending money to one of our contractors
and he sent me an invoice through Telegram,
I copy the wallet address, and it didn't copy the last letter,
the last symbol of the wallet.
So we lost some, you know, like, monthly salary, let's say.
Because, like, there was no notification, no any alarm about it,
there's no such a wallet, you know, like, anything like this.
And for agent, and right now our team even started
ideating like how we can implement it in our protocol, like on our platform, it could be
so easy, you know, like just to verify any activities that you're going to do as a user
before you're approving this or this transaction. Because it's very easy to track, like, does this
wallet exist? Is this wallet maybe marked as a suspicious wallet?
Or maybe a wallet that was involved in some
I don't know like money laundering or maybe scamming projects or whatever else
Is that a wallet that you interact often, you know, like and many other things that can be so much helpful for average user
so much helpful for average user just to improve his,
just to improve his
like just to protect by the end of the day, this user,
because like in crypto, we have really, really bad UX
and like high level of responsibility of the user
about any type of fuck ups.
And like, I'm in crypto like since 2017
and I did this mistake and I was like, what a fuck?
Like, how could I made this even mistake you
know like but it was after that I recorded the video and like I shared with
my team and everyone told like it's literally impossible to track it because
you just copy it it says it everything is copied and like by the end of the day
just one symbol was missed so I think like a lot of improvement could be on on
a user experience which gonna be reimagined, that we're going
to be experiencing and interacting between wallets, between other users, between other people,
just in super seamless and super easy way, because agents are going to be just abstracting a lot of
this tech overhead functionality. And on the other side, like how AI is changing and where I see
practical changes.
So we started two and a half years ago with developer support automation for many, many top tier protocols.
And it was like pretty good.
Right now we have more than 120 integrations with many protocols. And the core point is that development of applications never was so easy as right now so it's like if you
like ultimate language like a technical language skill is english language you know like you just
need to to write right prompts you need to understand what you want to build and you're
going to have it on any protocol on any ecosystem on anything with any smart contracts you need
and i think that's going to be a big unlock of application layer
for all the protocols and all the crypto world in general.
Thank you so much, Nick.
I mean, you're definitely working on very good problem-solving use cases
and kudos for that.
I just want to stay conscious regarding the time.
I will give last question to the community to ask.
The goal for this AMA that we run is not just to, you know,
bring the attention to the project, but also create like a mixer where,
you know, the big companies who already achieved some sort of success
and then the builders who already achieved some success
could share and collaborate with the younger startups and peoples who join.
So the last one questions for the community and you will definitely should follow Sint
and keep an eye on the next events and updates from us because we will be running AMAs on
a frequent basis.
So it's going to be every week and we will be bringing super cool speakers
all the time again kudos to everyone one last question and i will open it up
yeah so i just wanted to know how does fetch.ai didn't we regulate your challenges
against artificial intelligence as a player didn't really
can you go ahead and ask that question again, please?
I didn't quite catch it.
Yeah.
How is Fetch.ai dealing with regulatory challenges against artificial intelligence?
Well, the AI space isn't fully regulated yet,
so we're just monitoring the development of those regulations as they come in place.
We abide by every trad
regulation that exists as it relates to data protection if so if those apply or if they're
needed so broadly speaking we kind of have uh we obtain clarity through expert opinions and then
we kind of instantiate best practices that's the top level answer and then it's an emerging
technology right and it's often the case that the tech kind of speeds up.
So we're monitoring regulations as they emerge
and kind of adapting to the changing regulatory landscape.
I think that's really where I'd cap it off.
I appreciate it.
Appreciate it.
Thank you so much.
Thank you, GT Protocol.
Thank you, Fetch AI.
Thank you, Singularity.
Thank you, Sophiaverse, Liquify, Trend Dance, OpenX AI, Assisted, Strawberry AI, Glades AI,
and all other speakers.
Keep an eye on our community.
Keep an eye on our socials.
Please follow everyone who's been speaking this AMA.
Looking forward to build cool things together and chat with you all very soon.
Big pleasure.
Thank you so much again for joining.
Cheers.
Have a wonderful day.
Thank you, Elias.
Thank you, guys.
It was nice to meet you all
and speak about AI trends
and talking to 2049.
I wish everyone success.
Yeah.
Yeah, thank you.
Likewise.
Likewise.
Cheers, guys.
Bye-bye. Bye- Likewise. Cheers, guys. Bye-bye.
Bye, everyone.
Host
GT Protocol
from
GT-Protocol
