In this podcast, we invited Andrej, CEO and co-founder of Wynd Labs, to share with us his and his team’s experience in developing Grass network and real-time contextual retrieval (LCR) technology. Andrej introduced his background in transitioning from doctoral research in applied mathematics to entrepreneurship, and deeply explored how LCR technology changes the way AI models obtain real-time data, breaking through the limitations of traditional AI that only relies on static data.
We also talked about how the Grass network ensures data accuracy and transparency, and discussed the protections for user privacy in a decentralized network. Andrej detailed the technical and ethical advantages of the Grass network over traditional centralized platforms, and its wide range of applications in multiple industries — from e-commerce to financial market predictions. In addition, Andrej revealed Grass’ development plans in the coming months, including an upcoming Android app, new hardware devices, and incentives for contributors.
Audio transcription is done by GPT and may contain errors. Listen to the full podcast:
Microcosm: https://www.xiaoyuzhoufm.com/episodes/66f7e5da6c7f817786b5a762
YouTube: https://youtu.be/rFU9zOeWt30
Andrej's background and project overview
Prince：
Today we are very happy to have Andrej, CEO and co-founder of Wynd Labs with us. Andrej, can you tell us a little bit about yourself and how you got involved in this project and what exactly this project is about?
Andrej：
My name is Andrej. Before joining this project, I had been pursuing a PhD in applied mathematics, focusing on computational physics. During the pandemic, I started a company focused on web crawler infrastructure, providing customers with large servers and data centers dedicated to web scraping. As the number of customers exceeded 1,000, they began to ask if they could scrape from home devices and the Internet. At that time, I learned that companies like Walmart were using ordinary users' devices to scrape the web by secretly implanting SDKs in some free apps, which made me feel ethically uneasy.
Later, I met my two current co-founders and we decided to solve this problem together. We did not expect that with the rapid development of AI, this field would explode so rapidly.
Since we started, the project has grown tenfold. Grass is a network that anyone can join with one click, and after joining, users are basically renting out their own resources, such as CPU and bandwidth, to scrape public web information at scale.
Explaining Live Context Retrieval (LCR)
Prince：
Some of our listeners may not have an engineering or computing background and may not be familiar with AI technology. Can you explain your core technology — Live Context Retrieval (LCR) — to us in a simple way? Also, how does this technology change the way AI models interact with data? Can you compare it with traditional data training methods?
Andrej：
To explain LCR, imagine that when you are chatting with someone, if you mention a topic that they are not familiar with, maybe they haven’t kept up with the latest news. At this time, instead of giving you outdated information directly, they will first check the latest details and then give you a more accurate answer. The principle of LCR is similar, but it serves AI models.
Simply put, LCR enables AI models to access and use real-time information from external sources (such as the Internet) when generating answers or making decisions. This means that the AI model is no longer limited to what it learned when it was initially trained. Usually, when training a model, a large amount of static data is provided, so when you ask something that happened today or recently, it may not be able to answer. With LCR, it is equivalent to connecting an engine to the AI model, which allows it to access the Internet in real time and answer questions based on the latest information. The entire system is run by millions of nodes, and LCR operates on these nodes.
Comparison with Traditional AI Models
Prince：
Yes, I know that models like OpenAI do not have instant access to online information. They have manual intervention in the training process, and the feedback to the model will be re-labeled. So I want to know, does your system also have similar manual intervention, or is it fully automated?
Andrej：
Yes, there is absolutely no human intervention or human tampering in the Grass network. It is completely automated and everything is 100% transparently verified on the public ledger. This also ensures that all responses generated through the Grass network are completely unbiased and cannot be tampered with.
Information integrity and privacy issues
Prince：
Are you worried about the accuracy of information on the Internet? For example, AI may misunderstand some information, or give inappropriate answers, or even involve violent or nonsensical content. Has your team taken measures to prevent these problems?
Andrej：
Yes, every user of the Grass network must go through a KYB (Know Your Business) process and also adhere to certain compliance requirements. “Users” here refer to the customers on the other end, i.e. the ones running the AI models.
As for the data that AI models now access, they can currently only use search engine results. However, search engine results are full of ads, and while the content on the first page may be effective for humans, the quality of this content is poor for large language models (LLMs). Some research by OpenAI shows that in many cases, 40% of query results actually only use 1% of the data, indicating that a lot of data is useless.
The highlight of LCR is that since all nodes are verifying each data request in real time, we can ensure that the data returned to the AI model is not optimized for other purposes, but judged based on knowledge and semantic similarity. This is also an important difference of the Grass network.
Prince：
That's a really great point. Let's talk a little more about LCR. How does LCR handle privacy issues when acquiring real-time data? Are there any risks to user privacy?
Andrej：
Not at all. Grass nodes do not access your data or browsing history, the nodes simply utilize some CPU and bandwidth. The data being accessed is actually publicly available on the Internet and does not involve data on the user's personal computer. The user's device simply acts as a node to pass network requests. So there is no need to worry about personal privacy at all. For further assurance, we have been audited by three different organizations, one of which has been publicly available on our website. In addition, we have completed a security review with the Apple team, who are leaders in this field.
The impact of LCR in various industries
Prince：
Could you please talk more about the use cases of real-time data? How can LCR help industries outside of crypto, such as predictive models in financial markets, or other practical examples?
Andrej：
From a real-time data access perspective, every Fortune 500 company will use AI in some way in the next few years, and every application scenario will require access to the latest information. For example, an airline needs to understand the dynamics of the supply chain in real time and collect data from around the world to make decisions.
E-commerce companies like Target or Amazon need to crawl competitor websites in real time to understand their price changes. For example, Costco knows the prices of all products on Amazon every day. In the future, LLMs and AI will need tools like LCR to help them do these tasks efficiently.
The applications of LCR are very broad and not limited to a specific field. In the next few years, all industries using AI will need some form of real-time data access. Predictive financial market analysis is also an important application area for LCR. In fact, some top hedge funds have contacted us and want to participate in the closed beta test of LCR.
Grass's Competitive Advantage
Prince：
Yes, I see that there are many teams doing similar things now. What do you think is the biggest advantage of Grass compared to other competitors?
Andrej：
In the crypto space, I don’t think there’s any other solution that can do what we’re doing.
Prince：
So are there any competitors in the encryption field at present?
Andrej：
In crypto? Not yet that I know of. I expect some protocols will take notice of what Grass is doing and try to emulate it, but I haven’t seen any success so far.
Outside of crypto, our biggest competitors are two companies that crawl the entire internet. I can’t name them, but I’m sure you can guess.
Grass has two main advantages: First, on the ethical level, Grass is owned by users, all decisions are made by users, it is completely decentralized, and all content is recorded on an unchangeable public ledger that everyone can verify by themselves without relying on centralized companies.
Secondly, on the technical level, Grass can do what other companies cannot do because it is supported by millions of nodes and can crawl website information faster and more fairly. In addition, Grass has lower operating costs because it does not need to bear huge operating expenses like centralized companies. We rank pages based on the value of information to LLM, not based on advertising value, which makes us more efficient than other solutions.
Incentives for Contributors
Prince：
So what incentives do you offer for contributors to the network? How are users encouraged to participate?
Andrej：
That's a good question. Right now the network is completely passive, users only need to install a node, and the system will automatically do everything in the background. Although the application is passive, our community is very active, so we are also considering how to get users more actively involved. We will soon launch more consumer-oriented applications, the first one will use LCR, so everyone can look forward to it.
Regarding incentives, so far we have focused primarily on building the network and ensuring it is robust. Having millions of users join early on is amazing because it allows us to do large-scale stress testing and ensure the network is functioning properly. In return, early adopters will receive airdrops. To be honest, though, airdrops are not the main reason for people to download Grass, but more of a reward for early supporters.
In the long term, what we are really interested in is enabling bandwidth monetization. For example, when a large company uses LCR, if your bandwidth is used to query the data that company needs, you will be rewarded for it. This provides a passive source of income for users who simply run the system in the background without having to do anything extra.
Upcoming Developments
Prince：
Can you elaborate a little more on the rewards that users can expect? Are there any developments or major milestones you can share with us that we can expect in the next few months or year?
Andrej：
Absolutely. Next up is the Android app for Grass. A lot of you may have heard about the Saga app, which is the Solana phone. The Android app will be available after the network is live. We are also working on a consumer hardware device. The first product using LCR will be available soon, and users of Grass will have the opportunity to participate in testing. So if you already have Grass installed, you can experience this new product.
In addition, you can continue to pay attention to the roadmap we mentioned, such as the progress of upcoming hardware devices and other applications.