David Chen（GEODNET）：打造實體 AI 的定位基礎設施

0:00GEODNET 最新進展與全球規模

0:00大家好，我是 GEODNET 共同創辦人 David Chen。今天很高興在這裡跟大家分享我們專案的最新進展，以及 GEODNET 如何成為實體 AI 的基礎設施。首先，讓我快速更新一下我們目前的狀況。My name is David Chen, co-founder of GEODNET. Today, I'm very glad to give you some update of our project, the recent progress, and also our recent development, how GEODNET could be the rails for physical AI. So let me give you some quick update of our status.

0:32這是我們最新的 GEODNET RTK 網路。我們現在是全球最大的 RTK 網路，擁有 22,000 個基地台，遍佈全球，特別是歐洲、美國和許多西方國家（不包括中國和俄羅斯）。So this is the latest, our GEODNET network, the RTK network. We are officially the world's largest RTK network. It consists of 22,000 base stations all over the world, especially covering Europe, United States, and a lot of Western countries, while excluding China and Russia.

0:58要達到這個規模其實不容易。我們有 Web2 的競爭對手，例如美國最大的 Trimble 公司，這是一家市值數十億美元的上市公司。他們花了數億美元、二十年時間，才部署了約 5,000 個基地台。我們則採用區塊鏈去中心化方式，讓每個基地台擁有者可以在家裡用現有網路和電力部署並維護基地台，大幅降低成本。這也是為什麼我們四年內就能做到比第二大競爭對手大四倍。So to reach this state is not easy. We have Web2 competitors already there. So the biggest one in the United States is called a company called Trimble. It's a public company worth tens of billion dollars. They have spent hundreds million of dollars trying to deploy the similar network as we do, but they spent the last, you know, past 20 years, and they have a network about 5,000 base stations. And we approach, we take a different approach. And up to date, we are four times bigger than the public company. How do we do that? We are using blockchain, so we are doing a decentralized way. So every base station owners will deploy the station at home using existing network, electricity, and maintain the base station. Which tremendously save the cost. So that's why within four years, we can deploy four times bigger than the second competitor.

2:10極致精準定位與應用場景

2:10我們的網路可以提供高達 2 公分的 XYZ 定位精度，這比一般 GPS 準確 100 倍。這種高精度定位對自駕車、無人農機、精準農業等應用非常重要。現在很多農場的拖拉機已經是無人駕駛，24 小時自動作業，但要做到這點，定位精度必須優於 10 公分，一般 GPS 根本不夠力。我們的網路能做到 2 公分，讓這些應用成為可能。And second, our network can deliver two centimeter accuracy, X, Y, Z. That's about 100 times better than regular GPS. So because we can provide such an accurate precision network, it could be really useful for a lot of applications, like autonomy and automotive, and also precision agriculture. So a lot of the recent tractors running in farm field, they are already unmanned. So imagine, you know, those tractors running in the farm field 24 hours a day, 7 days a week, with a precision that can make it do automatically. But you need to have the accuracy up to at least better than 10 centimeters. The regular GPS won't work. But because our network can deliver two centimeter accuracy. They make it totally possible.

3:16再來像機器人、無人機，如果你要精準地從 A 點送到 B 點，也必須用我們的網路。還有自動割草機、VR 物件定位、HD 地圖等，都需要 RTK 精準定位。GNSS 也能用於高精地圖繪製。And also think about the robots and drones. If you want to deliver from one place to another precisely, you need to use our network. So let me give some real use case. This is a quick introduction. What's the difference between RTK and regular GPS? Another use case for our network for autonomous lawnmower. To get this precise positioning, you need RTK. And also for VR, if you're not the objects right in the place, you need a better positioning service. And also, GNSS can be used for HD mapping.

4:17真實世界營收與企業客戶

4:17因為 GEODNET 解決了真實世界的問題，所以我們有真實世界的營收。GEODNET 跟其他加密專案最大不同，就是我們的收入來自企業客戶，而不是賣硬體。到目前為止，我們已經有超過 165 家付費企業客戶，包括汽車、精準農業、機器人等領域，這些公司需要可靠且高精度的定位基礎設施來支援他們的產品和營運。Okay, because GEODNET solves real-world problems, it generates real-world revenue. What makes GEODNET different from a crypto project is we are getting revenue from the real world, not from hardware sales, but actually real business paying us. It's totally, you know, different from other crypto projects. So up to date, we have more than 165 paid enterprise customers, including like automotive, precision ag, robotics, and there are companies using GEODNET because they need a reliable, high-precision positioning infrastructure for real products and operations.

5:19目前我們的年化營收（ARR）已經突破 1,000 萬美元。這證明我們不只是蓋去中心化基礎設施，而是真的對企業有用，能產生實質價值。As a result, GEODNET has now reached more than 10 million USD revenue ARR from the real businesses. For us, this is important because that proof of GEODNET can be really useful. Not building the decentralized infrastructure, but it's useful for real businesses.

5:45代幣經濟模型與通縮機制

5:45所以，因為我們能夠從真正的企業獲得實際的收入。我們的代幣經濟模型是將這些收入的80%用來從公開市場回購代幣並永久銷毀。這是一個承諾。所有的交易和代幣銷毀都可以在鏈上被看見並驗證。所以這裡有兩張圖表，你可以在Blockworks網站上找到。左邊那張圖顯示每週銷毀的代幣金額（以美元計算）。你可以看到，過去幾年我們的收入成長，以及我們的銷毀成長。右邊那張圖顯示的是銷毀率。銷毀率的意思是以美元價值計算的代幣銷毀量。So because we can getting real revenue from real businesses. And our tokenomics is we are using 80% of those revenue to buy back tokens from open market and burn them permanently. This is a promise. The transaction and token burns are visible and verifiable on chain. So there are two charts. Which you can find in Blockworks website. The left one shows the weekly burn tokens in USD. You can see, you know, in the past several years, our revenue growth and also our burn growth as well. The right chart shows the burn rate. Burn rate meaning the token burned in USD value.

6:41相較於發給礦工鑄造的代幣，我們已經達到100%，這代表我們燃燒的代幣數量與我們鑄造的數量相同，甚至更多。所以這帶來了一個非常獨特的特點，就是我們正朝向通縮型的代幣經濟模型發展。Versus the token minted to the miners, we are reaching 100%, which meaning we are burning at the same level or even more than the token we minted. So that makes a very unique feature, like we are towards a deflationary tokenomics.

7:04總結一下，GEODNET 的飛輪是這樣運作的：我們在全球部署 22,000 個基地台，產生有價值的數據，企業付費使用，80% 營收回購銷毀代幣，形成正向循環，讓 GEODNET 成為極具吸引力的加密專案。So this is our summary of GEODNET tokenomics and how the flying wheel works. So we have stations deployed all over the world. We have 22,000 base stations right now, and we generate real useful data for real businesses, and they pay us. We get revenue from the real world. And we use 80% of the revenue to buy our tokens and burn them. And that makes a very positive flywheel to make GEODNET a very attractive crypto project.

7:39AI 時代下的 GEODNET 新定位

7:39那麼，GEODNET 下一步是什麼？在 AI 時代，我們有兩大獨特能力：第一，提供高達 2 公分的精準定位，這對 Web2 應用已經很有價值；第二，GEODNET 網路能提供可信賴的定位服務。傳統 GPS 很容易被偽造，像中東或俄烏戰爭區域，GPS 訊號很容易被干擾或假冒，但 GEODNET 能用去中心化方式驗證裝置真實位置，這就是我們的 proof of location（位置證明）協議。Okay, so that's the current status. The question we need to answer, or the question we need to ask is, what's next for GEODNET in the AI era. So let's review the two unique capabilities GEODNET offers. So first, we offer precise location up to two centimeter XYZ. So that's very useful for Web2, and that gets us a lot of revenue. And second, GEODNET network provides a trustable position service. Unlike GPS, so GPS signal is very easy to be faked. You know, imagine, you know, in the Middle East conflicts, Russia-Ukraine conflicts, the GPS there can easily fake. But for GEODNET, we provide a way that positioning is reliable and trustworthy. So, GEODNET can prove you are where you are. And that's basically, it's the native feature of GEODNET. We call it proof of location.

8:54我們用全球 22,000 個基地台，提供 2 公分精度的定位，並能去中心化驗證裝置是否真的在聲稱的位置。這對於需要可信位置的加密專案或其他應用非常關鍵，GPS 做不到這點，GEODNET 可以。So we are using 22,000 base stations all over the world, with a precise location up to two centimeters. And you can prove the device is actually in the place where you claim you are, and we can do the decentralized proof. This is very critical because GPS cannot provide that trust level, but GEODNET can. We essentially create a layer for crypto project and any other projects, really, relies on the trustable position.

9:29TokenRun：連結 AI 與真實世界的任務平台

9:29我們正在開發一個名為 TokenRun 的應用程式。TokenRun 是我們用於 agentic AI 的一個應用程式。對於這些 agentic AI 來說，他們已經掌握了網路上的資料，基本上已經消化了 90% 的網路資料，但他們並沒有直接獲取現實世界、即時資料的能力。對他們來說，想要取得即時、現實世界的資料，以前是做不到的，直到現在才有可能。如果他們想要取得資料資訊、即時資料，他們可以透過 MCP、透過公開的 API，將任務發送到我們的 TokenRun 應用程式。TokenRun 會把這些任務分派給真正的人類代理人。So we are working on an application called TokenRun. So TokenRun is an application that we use for agentic AI. So for the agentic AI, they already know the internet data. They pretty much consume 90% of internet data already, but they do not have direct information for the real-world, real-time data. So for them to get real-time, real-world data, it's not available before until now. So if they want to get data info, real-time data, they can send a task to our TokenRun application through a MCP, through API, which is publicly available. And TokenRun will dispatch those tasks to real human agents.

10:31人類代理人會前往指定地點，拍攝一些照片，獲得他們的代幣，並透過GEODNET網路驗證他們的位置，然後將這些資料上傳到伺服器，接著再分派給AI。因此，因為GEODNET能夠用加密的方式證明位置，這些資料是可信的。當然，人類代理人會以加密代幣獲得報酬。而AI代理人則能取得即時資料，進行處理並持續進行有價值的資料分析。The human agents will go to the place location, take some photo, get their token, get their position verified by GEODNET network, and then upload those data to the server, and then be dispatched to the AI. So because GEODNET can prove the location in a crypto way, those data are trustable. Well, of course, the human agents will be paid with crypto tokens. And the AI agents will get the real-time data, and they can process and continue the valuable data analysis.

11:13這樣一來，agentic AI 就能跨越虛擬與實體的界線，進入 physical AI。舉例來說，AI 可以即時檢查店面人流、排隊長度、營業狀態，也能現場查驗建案進度、設備狀況、交通路況、停車位、事故、競爭對手活動、促銷、群眾活動等。這些都是 GEODNET 能協助實現的應用場景。Now the agentic AI can cross the boundary into the physical AI. Let's give you some use cases AI agents can do with neural network. So they can check the store venue conditions, like foot traffic, queue length, whether the store is open, real time. And second is on-site inspection. You can go to the, you know, check the construction progress, equipment conditions, property status, and also traffic and mobility, like what's the available of parking and road conditions, transactions, accidents, and local market intelligence, competitor activity, pricing promotion, and events and crowd verification. So there are many use cases that can be done with the help of GEODNET.

12:14TokenRun 與 Work for AI 功能上線

12:14目前 TokenRun 已經在 Play Store 和 Seeker 手機上架，我們也是 Solana Seeker Store 上的第一名遊戲。現在我們加入了 Work for AI 功能，讓 AI 產生大量即時、真實世界的任務，這些任務會開放給 TokenRun 用戶，大家可以到現場蒐集資訊、用 GEODNET 驗證位置，然後上傳給 AI 使用。這種 Workflow AI 功能以前從未出現過，將會非常有價值。It currently is already available in Play Store and also in Seeker phone. We are already the top one game in Solana Seeker Store. Now we are adding another feature we call Work for AI. Try, you know, adopting the feature we described before. So the AI agent will generate a lot of real-time, real-world tasks. And the tasks will become available to the users of TokenRun users. They can go to places, get their information and location verified, and take real-time data and upload to be consumed by AI. So in that way, we can imagine the TokenRun Workflow AI will become a very unique and useful feature, which is not available before.

13:11GEODNET 與 Physical AI 的未來

13:11我們可以看到 agentic AI 正在進化成 physical AI，像 NVIDIA 等大公司也都朝這方向發展。我們相信 GEODNET 的 proof of location 協議是這個領域不可或缺的基礎，同時我們也提供 Workflow AI 平台，讓這一切成為現實。GEODNET 就是 physical AI 的基礎設施。歡迎大家追蹤我們的 X（推特）帳號 @GEODNET，敬請期待我們 TokenRun 應用的 Workflow AI 新功能即將上線！So let's look at this, you know, the agentic AI will evolve into physical AI, and that's like a lot of, like NVIDIA and other, you know, big-name companies, they are towards this way. And we believe GEODNET provides a unique way with proof location protocol, which is essential. And also we provide a platform, the workflow AI platform, to make it totally available. So that's our answer how a physical AI can be real into place. And GEODNET is a rails for the physical AI. So please follow us in X. Our X account is @GEODNET. And we have, stay tuned for our GEODNET, the TokenRun application workflow AI feature will be coming up very soon.