NEAR and Delphi Labs AI Accelerator Demo Day

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:00

   all right folks for the sake of time we're going to go ahead and get this party started I will pause our music Vibes here um well good morning or good afternoon good evening depending on where you're calling in from uh great to see everyone here my name is Jared Barnes uh I have the amazing opportunity to work with our incredible team at the near Foundation um and have spent the past three months working uh with the deli Labs team on building um and

2. 0:26

   growing our accelerator program which we are incredibly excited uh to share with you all today before we dive in we have some amazing demos and awesome awesome Founders that we um are really excited to to Showcase wanted to actually take a step back and provide some context on where this program came from and why we're doing this program but before I do that I will say this is actually a real

3. 0:50

   Zoom not a live stream so if you're unmuted if you could mute that would be greatly greatly appreciated um and if folks could help me uh here just mute folks so we're we're all set um for a good good experience but um really appreciate everyone joining to give you all some context right on where this program came from six months ago us at the near Foundation we asked ourselves how do we go build the most compelling

4. 1:15

   ecosystem at the intersection of AI and web 3 I think earlier this year there was probably still some questions is decentralized AI even going to be a thing I think the cat out is out of the bag uh for sure and uh we said okay if we're going to go build this ecosystem how do we go work with the best Partners in order to do that um we entered an amazing partnership um with uh Kevin and

5. 1:39

   the deli Labs team who turn to here in a moment um but I've been really really excited to not only um go find some of the best teams in the space uh but really work with them to uh grow and scale their products and projects which um we're going to unveil here in a moment so that's kind of it for me background framing but Kevin I'll turn it to you for some opening remarks and we'll we'll get right into

6. 2:03

   demos yeah thanks Jared I'll keep it quick so we can get to the alpha just to give everyone a little bit of background on Deli Labs we're a team of Builders we incubate and accelerate crypto projects we're part of the deli family you know along with Deli digital for research and Consulting and of course Deli Ventures and across Deli we've gone really deep into this intersection of crypto and AI I believe we've made about 18 Investments now out of Ventures we've covered the space pretty deeply through

7. 2:34

   reports and through podcasts but just getting to work with a near team and these eight projects that you'll hear about today has been an absolute pleasure couldn't be more excited about the talent on these teams and the products are developing and we really hope you feel the same fantastic Kevin thank you and the rest of the deli Labs team thank you um folks we're going to Dive Right In to these demos as Kevin mentioned we have

8. 2:59

   eight teams um and having been in the seats of an investor listening to backto back pitches uh we understand that actually can be a lot so we're going to take a short break in uh in between after the first four pitches and then um just so you have some context on our Logistics and Experience run of show here again this is a private Zoom specifically for everyone on this line after all the demos uh have been played it'll be about 40 minutes total um we're

9. 3:24

   going to have some breakout rooms where you will be able to connect with Founders individually and then for investors who would like to follow up um we'll be sharing some links um in order for you all to follow up and have access to all of the information um from Founders uh who uh will demo and present today as well hi I'm Roger Ying the co-founder of defusion AI where we're creating the data Hive for the agentic

10. 4:01

    web AI advancement is facing the data wall problem currently gpts and LMS are only mainly trained on the public web which is 4% of the internet that's already fully scraped for AI to advance it needs to reach the next 96% of private data and accessible to search engines major tech companies know this and that's why they're spending hundreds of millions of dollars to license data Enterprise that we trusted are now profiting of your personal and

11. 4:33

    professional data for AI training purposes which ironically will be destroying jobs and centralizing power to these major tech companies as we forecast AI agents to reach over 100 billion population normal model Builders struggle to find quality data to feed their agents We Believe data should come to you defusion AI allows you to attract data and information you want on demand how does this work operators

12. 5:04

    create subnets and set rubrics for data and information they want for example important companies in the semiconductor industry the protocol propagates your needs throughout the network which incentivizes both public and private contributors to submit data our defusion AI fine-tune models filter validate label and rag optimize the data in a vector DB storage for your Subnet in short we help you turn raw diesel data into high quality grade jet fuel for your AI

13. 5:44

    purposes now let's see what it takes to make a subnet you can specify relevant topics control factual or opinion based the data you want is and even write a rubric in your own words that the AI will use as a filter for the submitted data and information now let's try submit a file say a report from Nvidia the AI will score the submission based on your rubric label the data according to the relevant subtopics and don't forget that this

14. 6:14

    information is US of War to so here we can create an article based which will have the citations to the private data in the subnet so you can control the quality furthermore we will be adding subnet stats unique incentiv settings and knowledge graph visualizations for each subnet we did a MVP PO for web 3 topics where we received over 100 GB of text uploads our AI only allowed 35% to be filtered through and now we have over 38

15. 6:48

    GB of pure Text data which is already larger than Wikipedia we are processing over 50,000 data submissions per month only limited by our GPU capacity and we signed over over 40 crypto security related firms who want access to this data for their research purposes in our go to market we plan to launch together with Vanna where we're launching a DLP and have already secured a Bens of sponsor for our subnet allowing us to generate Revenue in tow and Vana tokens our main n subnets will

16. 7:19

    be able to host or create data feeds for other web3 Ai and data market places like koi public AI assister open Ledger and more being in the near ecosystem we plan to make use of chain abstraction with a goal to reach a much larger web 2 Enterprise market for our data services where we're already working with Stanford AI research and have other Enterprise interested parties leveraging

17. 7:45

    what we believe is a massive Arbitrage opportunity between web 3 and web 2 data services for AI we are planning a node license sale in January and February working closely with ather and in talks with impossible Finance test net in February TG in April and plan May net in May where Ai fintech and web3 Veterans leading a team of five developers I've grown a fintech to $300 million of AUM in China before which exited in 2018

18. 8:16

    Patrick myo has designed AR alos to detect blockchain fraud and brought products from0 to one the rest of the team has experienced at rip on binance and we're joined by great team advisers including the co-found of two Stanford phds McKenzie's ex AI lead as well as crypto hedge fund experience we're raising a strategic round now with 500,000 in soft commits and 1 million left to raise to be spent mainly on product and

19. 8:45

    marketing if interested welcome to reach out via email or telegram at r918 all right so our first team defusion we are really really excited about what they're building particularly in the data space as you all just heard from that pitch um if you're interested in meeting with Roger and the team we'll have a breakout room with them at the end um so thank you Roger and folks for uh that pitch

20. 9:10

    really really excited quick reminder before we go to our next pitch if you could just mute yourself um uh so we don't hear any background noise while the pitches are going on thanks folks we're going to go to our next team hi my name is Bob co-founder and CE of quest FL what we building is the orchestration layer for the multi-agent economy by allowing Builders to build monetize and tokenize a swamps so these swamps will be able to take actions and earn rewards autonomously on chain let

21. 9:48

    me show you how the product works in this demo so since we are talking about multi-agent then what kind of AI agents that we have you can simply click on the agent Tab and there are a bunch of agents already available in the marketplace some of the agents are more data focused like Hacker News GitHub OK link YouTube these are the agents can get data from specific sources with a specific knowledge another type of agents are action-taking ones like

22. 10:14

    Google Calendar telegram Gmail notion they can actually take actions on users behalf and if you click through all these agents you'll realize not only they have tools basically as skills what basically what things they can do templates basically how they can work with with other agents but also they all have a wallet address as well basically these are how these AI agents can get

23. 10:35

    paid on chain so let me show you how to use the multi agent SW in this case so there are two ways to use this one is you can simply describe what you want to do by creating a new I agent swamp on the spot the second way you can do is you can simply explore the existing aent swamp and then just get started so for example you can just get started with this which is a Twitter engagement swamp and what's happening is within this

24. 11:01

    there are some pre- defined AI agents available so in this case we already have a schedule agent for scheduling image generation agent for image creation Twitter agent for Twitter posting okay link agent for getting auction data and uh aggregated web three information for Co desk news basically so how you can use this you can simply describe natural language what you want to do and then we'll figure out okay in

25. 11:25

    order to do so how can we use all the AI agents within this agent group in order to take the actions for users so it takes a little bit of time to get started and what's happening is you can simply describe the user intent and as a orchestration protocol we will figure out okay what are the required AI agents in this case you want to do so how to prioritize them how to create a workflow

26. 11:51

    for them how to create a chain of thoughts or graph of thoughts in order to take actions for you and what are the missing pieces in order to accomplish this tax so so in this case because you're are creating a Twitter post about near protocol with a cat image thing the necessary step is we will require you to authorize us your Twitter account so what's happening is you can simply authorize right here and then confirm

27. 12:13

    now what's going to happen is since our orchestration protocol be able to understand okay you have all the missing pieces we have the authorizations to take actions what else are missing if there's nothing mything then we can simply just get started one by one by organizing these AI agents to take actions for you autonomously so what's going to happen here is you will see different agents will just start talking one by one and

28. 12:37

    they'll be able to figure out a workflow and get the tasks done for you so it takes a little bit time to get started and as you can see the aggregated web 3 agent is getting the web 3 data the image generation agent is getting the image created the Twitter agent is drafting a tweets with the image with the Twitter content and as you can see a Twitter post has successfully been created and will be posted soon with content with you know image and also with you know

29. 13:17

    hashtags as well and what's happening behind the scene is you will see that you'll be able to tell there is already a task getting created and all these task execution data will be work work record for these AI agents on chain and the next step will be you can simply click through agentic dashboard and within this dashboard you will see a bunch of AI agents and how they are working and and how much they're getting

30. 13:42

    paid on chain so how you're looking at this is there will be a merchant basically whether it's Quest flow or third parties using the service the user the orchestration protocol and the uh basically the agents getting paid if you click through this you'll realize these agents are getting paid on chain right now and you can also click through the hash just to see they are getting paid launching as

31. 14:08

    well yeah and there are more there are way there are more features there are more features that they can use within cost flow you can create new things by uploading knowledges you can also explore other agent swamps that's available feel free to give it a try and let us know what you need thank you amazing we are really really excited about what quest flow is building particularly uh a very user-friendly

32. 14:32

    interface for multi-agent orchestration um so highly recommend and encourage you all reaching out to Bob and the quest flow team as well we'll keep it moving we'll go to our next demo um we'll do a couple more and then take a quick quick break hey all really excited to talk to you about our company magic magic on Rams private data for consumers to enable the next generation of models

33. 15:01

    agents and applications our vision is to power an open and fair AI ecosystem where both data contributors and developers get fairly compensated before we dive into the product let me tell you a little bit about the founders backgrounds so I'm David I'm our CTO I have a BS and Ms from Stanford studing Cs and AI I worked on LM fine tuning at the Stanford NLP group and I also worked at many AI industry teams like repet on

34. 15:23

    their co-pilot and Salesforce on the Einstein LM product chain is my co-founder is the CEO he studied CS in economics at Stanford he worked at Federated learning under Professor Andrew ining at the Stanford AI lab where he was first introduced to private model training and he also worked at scale AI on training data pipelines so Chay and I met at our dorm at Stanford we initially went through YC actually launching an alternative payments product where users could pay via Bank

35. 15:48

    by connecting via plaid we aggregated 80 million data points with that product and facilitated 50 million total tpv and we shifted to this current space because of two primary reasons one hedge Fones began approaching us to access the transaction data we had through different bank accounts and it became actually one of our largest revenue streams it may how realize how valuable private data is at the same time a lot of our friends working at YC starps and AI all said the same thing they need

36. 16:13

    access to private data to improve their models to be competitive because all everyone is essentially using the same public data and so it's pretty clear to us that the lack of private data for AI is the one of the biggest problems right now and given our expertise that's what we should be working on so for some context the era of public data is coming to an end and we are entering the era of private data why because we've exhausted essentially all public data hence the data wall to put

37. 16:38

    into numbers meta's latest llama 3.1 model is built on top of 14 trillion tokens which is essentially the extent of all public data and the next generation of models won't be 10x smart without access to private data private data is the critical resource and demand has always outpa supply for private data the problem is you can't access private data very easily it's locked behind authentication flows personal devices Enterprise databases to access private

38. 17:02

    data you need to contribute private data the current infrastructure is not welld designed to handle private data because once data is made public it can be endlessly copied and shared which renders it useless from a monetary standpoint so there's where our platform comes into play our solution to this is a decentralized private data platform for AI so what does it look like the user flow is this first we aggregate

39. 17:25

    user data users contribute their private data in our consumer and choose their data privacy level from transparent to highly secure we're very comfortable with this because we've done it before we previously aggregated 80 million data points and we know that getting users contribute their data is about rewards and convenience private data is then encrypted and stored offchain while permissions are managed onchain and then we score and validate the data with

40. 17:47

    novel scoring functions within tees and NPCs and then finally the heart of our platform private model training private model training allows developers to train models on private data sets without unex exposing the underlying raw data and this solves the main issue with private data once it's public it can be endlessly copied and shared because private model training never exposes the raw data data Integrity can be preserved which ensures Fair contribution com

41. 18:12

    compensation and access our vision essentially is a private data enabled developer ecosystem where developers train on the world's largest network of user owned private data so developers will essentially use the private data sets we aggregate to build smaller expert models why specialized applications instead of large language models well because private data sets are the best fit for models that are fine-tuned for specific tasks and domains in our view the future

42. 18:39

    of AI is like a barbell we'll have a few llms from open Ai and topic on one side and millions of expert models on the other side all powered by Magic and you know many of the popular crypto xai ideas around agents that make payments or apple intelligence for the masses is not possible without true expert models powered by private data so the premar platform as well is an AI native specific blockchain where onchain data scores and transactions are uh

43. 19:05

    recorded by the blockchain and a platform native token is used to access the data sets we have secured 34 Lois from some great YC aai startups and we have some great Partnerships as well with different crypto xai startups in the ecosystem and we recently also raised an oversubscribed round from some great VCS including Deli Ventures and we'd love to uh keep the conversation going thank you

44. 19:28

    you awesome well we are also excited about magic as well particularly what they're doing at the data layer in a really really compelling way we're going to go to our next pitch uh and then take a short break after that in 20123 Deep's total addressable Market hit $2.2 trillion now it's on track for 3.5 trillion by 2028 its six core categories compute wireless energy AI Services sensors and service Market places each approach

45. 20:08

    trillion doll scales but deepen current $45 billion market cap lags far behind the reason today's deepen networks are heavily centralized they rely on offchain orchestration and scheduling through AWS and most connected devices lack the gpus needed for larger AI models these constraints all hold deep in back meat periphery first vertically integrated AI infrastructure built to unleash deepens full potential our Stacks consist of four layers an AI workflow Suite where developers can

46. 20:43

    train find tune and optimize models for Edge specific inference making every device truly intelligent an edge infrastructure layer which enables these models to be infused with richer context aware capabilities a unified device layer which Aggregates and balances load across multiple deepen networks ensuring decentralized and efficient scaling and an evm compatible blockchain which enables Inc contract inference so smart contracts can evolve and improve over

47. 21:13

    time together these layers replace today's fragmentation with a seamless intelligent ecosystem berey makes decentralized Network smarter more scalable and fully verifiable unlocking the multi-trillion dollar promise of deepin Imagine optimizing supply and demand across all deepin Platforms in real time in a decentralized way our UniFi device layer integrated with Partners like i.net spheron and gradient Network enables efficient resource

48. 21:45

    sharing ritual will utilize this layer to power infernet nodes meanwhile our AI workflow Suite helps neillian deploy larger models by sharding them across secure enclaves for verifiable inference sentient zero gravity open gradient Talis and others leverage our compression stack to serve models more efficiently these collaborations demonstrate early signs of product Market fit establishing trust and momentum as we expand layer by layer with a top household names in

49. 22:20

    crypto ry's team brings deep Technical and Industry experience ADV Vite our CEO has 57 AI Publications is a former UCL lecturer and recently lectured at the London Business School Matthew our CDO has a PHD in computer science from Imperial College London specializing in resource optimization Kim our CPO is an imperial physics graduate and an early hirer at Fidelity blockchain team also a veteran of low latency Financial systems BC our

50. 22:54

    coo is an imperial alumnus with experience at Google Nvidia and Activision Blizzard now completing his MBA at MIT slowen we blend top tier research institutional rigor and practical knowhow to bring the future of AI to deep in we're looking to raise a seed round of $3 million we currently over 2 million in follow on interest this round the funding will enable us to deliver the periphery mainnet in 2025 and

51. 23:25

    provide an operational Runway of 24 months the future is peripheral history shows transformative Technologies need enabling infrastructure electricity needed a grid steam engines required Railways and the internet demanded personal computers for deep in to reshape our world it needs its own Catalyst periphery is that Foundation turning deep in from a fragmented ambition into a powerful

52. 23:55

    pervasive reality all right folks so you've heard from our first four teams with per Labs being the most recent one and we're really excited about the Unique Piece of infrastructure they're building specifically for model compression right in quantization of large models and making them usable um in everyday consumer devices all right folks very excited about the remainder of this cohort we're going to close out these demos thanks a lot to Del and for hosting me

53. 24:31

    today I'm Chi co-founder and CEO of kaii Kai is feing the coordination layer for community be AI enabling scalable access to data models and agents with robust ownership and promise AI is projected to add $2 trillion to the global economy in 5 years driving billions of specialized the models in agent-- driven applications while GPU investment have substantially AED the compute bonck high

54. 24:56

    quality data has become the next big major challenge for example GT4 TR on just a one py of data while JP Morgan alone holds over 150 pte highlighting untapped potential locked in non-public data silos across cor operations and consumers the challenge centralized the system limited user control and unclear ownership prevent Fair compensation and hinger collaboration between Ai buers and data creators kaii solves this with a decentralized layer enable in secure and

55. 25:29

    scalable access to data models and agents using purpose vied identity and attribution mechanisms we Empower businesses and individuals to own monetize and collaborate in building AI together now let's watch Kai in action kite aai Bridges the gap between data providers and AI Builders unlocking the potential of unique highquality data sets while ensuring security transparency and fair rewards why kite aai traditional data marketplaces are enough when data is sold providers lose

56. 26:01

    control it can be resold or used without consent diminishing its value and exclusivity providers also face undervaluation as AI Builders often profit far more from the data than its original price leaving providers under rewarded and for Builders the lack of quality assurance creates hesitation resulting in missed opportunities to collaborate kid changes this we've redefined data collaboration with two flexible pricing mechanisms designed to

57. 26:30

    protect both sides while enabling success Revenue sharing lets data providers share access not ownership developers can use the data without upfront fees and both intellectual properties remain protected no data or AI deliverables can be downloaded when AI applications generate Revenue both sides share in the success the one-time payment option allows developers to test data risk-free

58. 26:56

    data stays secure throughout the process and cannot be resold if the developer decides to publish the data provider receives an upfront payment whether you're a data provider or an AI Builder kite AI ensures security and fairness providers retain control over their data while opening new ways to monetize Builders gain risk-free access to Unique data sets allowing them to create confidently and publish on terms that work for both parties kid aai also

59. 27:25

    accelerates deployment and monetization Builders can deploy models at scale earning Revenue through API calls or direct interactions via our userfriendly interface providers benefit from transparent ownership tracking with every claim and reward accounted for let's take a closer look at the product here our query engine and AI Lake come into play it tracks data ownership with Precision ensuring both data providers and AI Builders receive Revenue at their claimed

60. 27:55

    price it also simplifies how Builders interact with data sets streamlining access and usage for seamless collaboration while it protects both data and AI deliverables offering Builders Peace of Mind kite aai isn't just a platform it's a new ecosystem for AI collaboration data providers can securely monetize their resources and connect with Innovative Partners while AI Builders can access exclusive data

61. 28:21

    sets scale their creations and monetize with ease together we're soaring to new heights in the future of AI our team brings deep Ai and Weare expertise I completed my PhD in machine learning at us Berkley with over a decade of AI experience Michael funder Scott former and Community chair Uber designed Ai and data infra Uber and Salesforce lay our business leader was the head of B and near Foundation backed by top investors like J CIS and samx

62. 28:53

    we're raising the next round with strong commitments from existing investors like GC we love to connect with potential investors in the coming weeks to learn more follow us on a at goai aai visit our website go kite. or scan the QR code thank you what a way to kick off our final four teams very excited about everything that is happening uh through kid they've built an incredible

63. 29:17

    Foundation of a great uh database and data indexing uh company in zabok and now really taking that to the next level um so very excited about what they're building we're going to keep things moving we'll go to our final three teams and to our next demo my name is Yang Tang I'm the CEO of qar labs qstar labs is an AI application company focused on building autonomous agents at scale we are at the dawn of AI agents 5 years from now we'll see billions of Agents transforming how we

64. 29:53

    interact in this world we're not watching this revolution we're defining it we have the platform Vision team and culture to be the leader in this space our influencer platform in Meda is not just another AI tool it's an intelligent ecosystem where personas learn adapt and grow we first became obsessed with AI and machine learning and our careers on Wall Street trafi was the first place where AI was deployed at scale the majority of capital markets trades electronically today a testament to the

65. 30:24

    power of AI over the past few years we have seen an explos in generative AI along with public interest we formed qar labs to bring our knowledge and IP from financial markets into Mass public adoption currently AI influencers are painfully basic or very experimental they're glorified GPT rappers that lose or shine after a few interactions while this is novel for many people that have never been exposed to AI it certainly will not retain users ouri agent framework is very different and we focus

66. 30:55

    on three areas of breakthrough personaliz custom knowledge base reinforcement learning engine and proprietary agent rails the first area of focus is the personalized Knowledge Graph we start with CD knowledge in Persona in an easy to use no code required manner this is built into a custom knowledge base using retrieval augmentation to generate a worldview and content strategy of each influencer we

67. 31:19

    link this to content generation to make short and long form content images and videos we have a number of tones tuned and every piece of content generat at is used for selflearning purposes either through human driven training or L into the while to gather data REI influencers have a wide range of capabilities including chat and can live on any platform this process allows for vastly improved content creation and engagement process the second area of focus for us

68. 31:47

    is using reinforcement learning for strategic Behavior an influencer may have a goal of Engagement or a CTA and the content generation and actions performed should differ this is the next level for agents using reward functions to guide behavior and exploit data patterns towards a strategic goal the last area of focus for us is the proprietary agent functionality rails if you ask GPT what the price of Bitcoin is it'll likely give you a random number

69. 32:14

    pull from somewhere for an influ for an influencer to be useful and perform tasks it needs to be domain aware and this is the long-term power of our platform influencers evolving into autonomous AI agents starting with a AI influencers as a fun and lowrisk way for users to learn engage and grow with AI an influencer can evolve into Endless Possibilities analyst Trader PM advisor

70. 32:37

    to name a few these functions are domain specific and require proprietary training and rails many of these tasks are quite repetitive and this is where applied AI adds an edge we're focused on three initial areas crypto prediction markets and financial markets this is the domain expertise of our team and all three are probabilistic Returns on Capital our go toet Market is to empower users to build their own influencers and

71. 33:00

    then to launch a token to engage and build their Community creating tokens further enhances participation experience of users and gives them ownership in their creation meanwhile we get users comfortable with AI for more agentic functionality build brand and mind share and collect personalized data we have a number of K's working with us helping guide the roll out Key to Our Success is the experience and culture of

72. 33:23

    our team we have strong backgrounds in industry and Academia I have launched the Quant firm using ml optimization to manage portfolios Carlos previously ran the E trading business at DB and BFA and launched the crypto trading firm we have experience as entrepreneurs building businesses within major organizations and managing Talent our AI bench is quite deep with over 40 years of combined experience in working with AI ml our chief researcher was a postto

73. 33:48

    Asser and let the Quan team for liquid rates at DB we have an experienced Dev team with deep infro engineering experience and a number of candidates lined up ready to scale with us we emphasize D and R&D in our culture a winning mindset and a cohesive team spirit our core team has a long and professional personal history together we're currently raising our seed round ahead of a TG planned for early q125 the AI Revolution isn't coming it's here we're not participating we're

74. 34:18

    defining it we have the vision product strategy and most importantly the team are you ready to shape the future with us thank you well with all the momentum around AI agents right now we are very excited about qar um and all that's happening with them and the opportunity space there as well so for sake of time we're going to keep moving um but really really appreciate qar in the pitch there so thank

75. 34:48

    you my name is Lars I co-founder and CTO of a company called Alman and to start off this demo I want to pull your attention to two trends that are currently happening um the first one is Automation in financial markets now this has been going on forever um uh but currently over 90% of all uh volume in financial markets is driven by automated strategic software applications is not by humans and the same is coming to to decentralized comps M Bots Arbitrage

76. 35:21

    Bots successful profitable Market making is today already all done by software applications so that's Trend number one and then Trend number two you you're all a well this but for AI I want to focus on the quantitative reasoning skills of uh AI these are getting a lot better there are specific benchmarks that track how good llms are at at quantitative reasoning and they're getting

77. 35:46

    increasingly good so these are two Trends and they come together to make something uh beautiful There's Something Beautiful at the intersection of these two and that is financial agents so that's what I want to talk to you about today so what are Financial agents there's a lot of Agents out there a lot of companies doing stuff with agent so let's define and let's put the bar for what is an agent so an agent is a

78. 36:11

    software application that can extract information from a changing environment um changing is important because markets change new depths has pop up um so uh it should be able to read from those um is able to reason about the information that it finds obviously this is in the context of this financial goal which in most cases is is to make money it is able to discover patterns learn those and utilize them and then finally it can

79. 36:44

    create a test and deploy strategies to capitalize on what it learns in the patters that it finds so essentially they do they can do Quant work now how how are they different from quants and how will they augment quants they can discover more opportunities than quants can they don't sleep they can take in as much information as they want at the same time so that's a big difference they can also come up with more strategies as these models are

80. 37:13

    getting better they're eventually going to surpass Quant in in how complex they can make them um they can capitalize faster and opportuni so their their Cycles just faster because their computers and at the end of the day they will also prefering than um what what quants can do if you have these these top three things going for you you're you're you're going to be better on so how will final ages change the Quant workflow this is a very famous Quant you might you might recognize them

81. 37:45

    so how they work is they analyze uh the markets to try and hunt for Alpha they build a strategy they test it to see if it actually works they optimize it for the current marketing conditions and then they deploy deploy strategy and monitor it and then go back to the drawing board agents will come in and assist squads and hold this uh steps so they will be the co-pilot initially soon after that as these uh models and the the workflow gets better the Quan will actually be the co-pilot

82. 38:17

    so the AG will take over and do most of the work and the qu will just verify it and then finally as the model gets so good it's not going to take a Quant to to do the co-piloting in work can be anyone really that's what elik is doing we are facilitating this transition in the financial agent Revolution and specifically for defi we've been doing so for the past two years and we built a def5 strategy automation platform we're currently in close Alpha we're going to release it in

83. 38:50

    public beta next month so very excited about that and in q1 2025 we will be bu be releasing these quantitative reasoning models that will act as agents and can do all of this on your behalf and essentially match and gr your wealth I have a short demo uh that I want to show you of what our platform looks like I know this is not enough to show you all the capabilities that it has we're super excited about launching this in the next month so if you want to help out please

84. 39:23

    apply I have a QR code that I can show you here so if you're a strategist sign up if you're a defi Builder reach out to me you can do that also using the QR code and if you're an investor or you want to build with us do the same so thanks for your attention and yeah have a great one