Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Grass Founder Andrej explains why more web scraping doesn’t mean better AI data. “Anybody these days can go and deploy a crawler and with enough compute, you can go and advertise that you’re crawling insane numbers of pages. The thing is, there’s a lot of useless stuff on the internet.”

Delphi Digital

1,606,857 subscribers

44,955 Aufrufe • vor 14 Tagen •via X (Twitter)

Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

8 Kommentare

Profilbild von Knersuz

Knersuzvor 14 Tagen

@0xdrej Why are they building on Solana and why is their X account so inactive? Otherwise impressive!

Profilbild von Zynta

Zyntavor 14 Tagen

@0xdrej The key is the discernment

Profilbild von Daniel Vargas

Daniel Vargasvor 14 Tagen

@0xdrej Eso y que sin un filtro de calidad para limpiar el ruido, terminas entrenando el modelo con basura que cuesta caro procesar.

Profilbild von Monolit

Monolitvor 14 Tagen

@0xdrej grass token's sitting at ~$485m mc on orca fyi

Profilbild von CatGod

CatGodvor 14 Tagen

@0xdrej Perhaps the GODS of DATA, in their infinite wisdom, choose to bask in fragrant incense of quality rather than a deluge of stale scrolls…

Profilbild von Omega

Omegavor 14 Tagen

@0xdrej Andrej Radonjic just made web scraping sound as boring as it probably is.

Profilbild von Raphaël Aɱbrosius Costeau

Raphaël Aɱbrosius Costeauvor 14 Tagen

@0xdrej Fuck grass and fuck drej scammer

Profilbild von Macro Bombastic

Macro Bombasticvor 14 Tagen

@0xdrej yeah scraping volume is a vanity metric tbh quality beats quantity every time

Ähnliche Videos

Grass founder Andrej explains how AI labs price training data. “We’ve got a specific benchmark that we want to improve a model on. What is the expected economic benefit of making the model better on that benchmark?”

Grass founder Andrej explains how AI labs price training data. “We’ve got a specific benchmark that we want to improve a model on. What is the expected economic benefit of making the model better on that benchmark?”

Delphi Digital

13,337 Aufrufe • vor 19 Tagen

Grass Founder Andrej explains why AI labs are racing against the clock. “The work that goes into taking huge amounts of data and turning it into what you want to actually train your model on is massive, and it takes time. And time is probably the most valuable resource right now in AI.”

Grass Founder Andrej explains why AI labs are racing against the clock. “The work that goes into taking huge amounts of data and turning it into what you want to actually train your model on is massive, and it takes time. And time is probably the most valuable resource right now in AI.”

Delphi Digital

10,979 Aufrufe • vor 17 Tagen

New Delphi Podcast with Andrej Radonjic Andrej, founder of Grass To a website, Grass looks like normal users browsing. Behind it is millions of opted in residential nodes and a data platform that has delivered exabytes of web data to the frontier labs We get into how AI flipped the hard part of search upside down. The index used to be the moat, and crawling was free. Now AI tooling makes the index cheap, while AI crawlers have pushed the web into blocking everything. The hard part is getting the page at all Training data is already a profitable business for Grass with most of the top labs and multiple Mag 7 companies as customers. Crawling for that data also means mapping the entire internet, which is what puts Grass in a position to build search on top of it. Contents API is live, Search is next, and multimodal grounding follows 00:00 Intro 00:43 From Nuclear Physics to Web Scraping 03:38 The Pandemic Signal Everyone Missed 07:31 The Smart TV Proxy Scandal 09:10 Why the Internet Is Blocking AI 11:56 Training Data vs Live Data 16:05 Mapping the Entire Internet 18:03 Delivering Exabytes to AI Labs 21:48 What AI Labs Really Pay For 24:43 How Robots Learn From Web Video 33:02 When Labs Change Direction 35:57 One API Call for Any Website 41:58 AI Flipped Search Upside Down 47:03 Can the Network Keep Up? 49:52 The Defense Tech of AI Data 52:58 Why Data Is the Alpha 57:08 Why AI Labs Won’t Buy Slop 59:17 The Metric That Matters Most 01:02:35 The $100 Billion Data Market 01:05:07 Who Actually Owns Grass 01:08:18 What a Real Moat Looks Like 01:13:16 What Happens Next Disclosure. Delphi Ventures holds $GRASS Full episode

New Delphi Podcast with Andrej Radonjic Andrej, founder of Grass To a website, Grass looks like normal users browsing. Behind it is millions of opted in residential nodes and a data platform that has delivered exabytes of web data to the frontier labs We get into how AI flipped the hard part of search upside down. The index used to be the moat, and crawling was free. Now AI tooling makes the index cheap, while AI crawlers have pushed the web into blocking everything. The hard part is getting the page at all Training data is already a profitable business for Grass with most of the top labs and multiple Mag 7 companies as customers. Crawling for that data also means mapping the entire internet, which is what puts Grass in a position to build search on top of it. Contents API is live, Search is next, and multimodal grounding follows 00:00 Intro 00:43 From Nuclear Physics to Web Scraping 03:38 The Pandemic Signal Everyone Missed 07:31 The Smart TV Proxy Scandal 09:10 Why the Internet Is Blocking AI 11:56 Training Data vs Live Data 16:05 Mapping the Entire Internet 18:03 Delivering Exabytes to AI Labs 21:48 What AI Labs Really Pay For 24:43 How Robots Learn From Web Video 33:02 When Labs Change Direction 35:57 One API Call for Any Website 41:58 AI Flipped Search Upside Down 47:03 Can the Network Keep Up? 49:52 The Defense Tech of AI Data 52:58 Why Data Is the Alpha 57:08 Why AI Labs Won’t Buy Slop 59:17 The Metric That Matters Most 01:02:35 The $100 Billion Data Market 01:05:07 Who Actually Owns Grass 01:08:18 What a Real Moat Looks Like 01:13:16 What Happens Next Disclosure. Delphi Ventures holds $GRASS Full episode

Yan Liberman

107,871 Aufrufe • vor 20 Tagen

🚨 THIS IS ACTUALLY INSANE Your AI agent can have access to the web. But if it can't reliably read what’s actually on the page, that access is almost useless. We looked at Firecrawl as the web layer for AI agents and the numbers are hard to ignore. The setup is simple: Give it a URL, search query, or website. Firecrawl handles the ugly part — crawling, scraping, rendering, extracting, and turning web content into something an AI model can actually use. The headline numbers: → 173,000+ GitHub stars → Search, scrape and interact with the web at scale → Supports web pages, PDFs, DOCX and other content → Structured data extraction for AI workflows → MCP support for connecting it directly to AI agents The workflow looks like this: Search → Scrape → Crawl → Extract → Feed the agent Three things stand out: 1. Scraping becomes an infrastructure layer Instead of maintaining your own pile of HTTP clients, parsers, browser automation and retry logic, you can treat web access as an API. 2. Agents get more than raw HTML The goal isn't just downloading a webpage. It's turning messy web content into clean context that an LLM can reason over. 3. The same layer works across different agent workflows Research agents. RAG pipelines. AI search. Competitive intelligence. Web-data extraction. The interesting shift: AI agents don't just need better models. They need better access to the information those models are supposed to reason about. Firecrawl is building that layer. Save this repo.

🚨 THIS IS ACTUALLY INSANE Your AI agent can have access to the web. But if it can't reliably read what’s actually on the page, that access is almost useless. We looked at Firecrawl as the web layer for AI agents and the numbers are hard to ignore. The setup is simple: Give it a URL, search query, or website. Firecrawl handles the ugly part — crawling, scraping, rendering, extracting, and turning web content into something an AI model can actually use. The headline numbers: → 173,000+ GitHub stars → Search, scrape and interact with the web at scale → Supports web pages, PDFs, DOCX and other content → Structured data extraction for AI workflows → MCP support for connecting it directly to AI agents The workflow looks like this: Search → Scrape → Crawl → Extract → Feed the agent Three things stand out: 1. Scraping becomes an infrastructure layer Instead of maintaining your own pile of HTTP clients, parsers, browser automation and retry logic, you can treat web access as an API. 2. Agents get more than raw HTML The goal isn't just downloading a webpage. It's turning messy web content into clean context that an LLM can reason over. 3. The same layer works across different agent workflows Research agents. RAG pipelines. AI search. Competitive intelligence. Web-data extraction. The interesting shift: AI agents don't just need better models. They need better access to the information those models are supposed to reason about. Firecrawl is building that layer. Save this repo.

Vikas gupta

20,121 Aufrufe • vor 2 Monaten

In the second episode of Scenius Studio's mini-series "The Use-Case", I sit down with Andrej Co-Founder of touch grass. Grass gives users the ability to earn ownership in the Grass network by supplying the protocol with their unused internet bandwidth for data scraping purposes (something that is already happening to most of us and we don’t get paid!). The grass protocol packages this scraped web data and sells it to AI companies who have insufficient data to further develop their models. With over 3 millions users and millions of annualized revenue, Grass is a real commercial business with a roadmap that makes it one of the most exciting projects at the intersection of crypto x AI and data. In this episode we discuss: ➔ Andrej’s background in physics, finance, and sports betting ➔ Big companies using your IP address without your knowledge or permission ➔ How the Grass protocol puts a toll booth on your internet bandwidth highway ➔ Packaging web scraped data and selling it to AI companies building Multi-Modal models ➔ Dynamics between the Grass Protocol and the labs entity developing Grass’ IP ➔ Protocol design decisions to ensure that all tokenholders (VCs, team, and community) are aligned ➔ Why Grass needed to be built on crypto rails to maximize its potential ➔ The future of LLMs and how they will search for context and information Hope you enjoy this episode of Scenius Studio's "The Use-Case". Links to listen in bio or below👇

In the second episode of Scenius Studio's mini-series "The Use-Case", I sit down with Andrej Co-Founder of touch grass. Grass gives users the ability to earn ownership in the Grass network by supplying the protocol with their unused internet bandwidth for data scraping purposes (something that is already happening to most of us and we don’t get paid!). The grass protocol packages this scraped web data and sells it to AI companies who have insufficient data to further develop their models. With over 3 millions users and millions of annualized revenue, Grass is a real commercial business with a roadmap that makes it one of the most exciting projects at the intersection of crypto x AI and data. In this episode we discuss: ➔ Andrej’s background in physics, finance, and sports betting ➔ Big companies using your IP address without your knowledge or permission ➔ How the Grass protocol puts a toll booth on your internet bandwidth highway ➔ Packaging web scraped data and selling it to AI companies building Multi-Modal models ➔ Dynamics between the Grass Protocol and the labs entity developing Grass’ IP ➔ Protocol design decisions to ensure that all tokenholders (VCs, team, and community) are aligned ➔ Why Grass needed to be built on crypto rails to maximize its potential ➔ The future of LLMs and how they will search for context and information Hope you enjoy this episode of Scenius Studio's "The Use-Case". Links to listen in bio or below👇

Ben Jacobs

22,468 Aufrufe • vor 1 Jahr

"When you win and you know that you can play a lot better, that is a good, good thing." Go inside the 49ers locker room following the win.

"When you win and you know that you can play a lot better, that is a good, good thing." Go inside the 49ers locker room following the win.

San Francisco 49ers

114,627 Aufrufe • vor 2 Jahren

Will touch grass become the data layer for AI? In this weeks podcast I sit down with the founder Andrej as we speak to the AI boom and necessary ingredients needed to power it: compute and data AI x Crypto x Data (Grass) We speak to: - 600,000 users in only seven months - Expanding into data cleaning and preparation - What are the current issues in the AI space right now, and how is Grass addressing these issues Youtube: Spotify:

Will touch grass become the data layer for AI? In this weeks podcast I sit down with the founder Andrej as we speak to the AI boom and necessary ingredients needed to power it: compute and data AI x Crypto x Data (Grass) We speak to: - 600,000 users in only seven months - Expanding into data cleaning and preparation - What are the current issues in the AI space right now, and how is Grass addressing these issues Youtube: Spotify:

Logan Jastremski

136,484 Aufrufe • vor 2 Jahren

Grass users contribute bandwidth, not their personal data “Anyone joining the Grass network isn't actually sharing, like, personal information or anything like that” “Ultimately, what they're contributing is just internet bandwidth” “When AI labs need to build huge training data sets to train their AI models, the most important source for them is actually the internet” Andrej clears up a common misconception around Grass. The network uses participants’ internet bandwidth to help access and collect public web data. Users provide the connection layer, while the web itself supplies the information used to build datasets. As more of the internet becomes difficult to access programmatically, distributed bandwidth becomes part of the infrastructure connecting public web data with AI.

Grass users contribute bandwidth, not their personal data “Anyone joining the Grass network isn't actually sharing, like, personal information or anything like that” “Ultimately, what they're contributing is just internet bandwidth” “When AI labs need to build huge training data sets to train their AI models, the most important source for them is actually the internet” Andrej clears up a common misconception around Grass. The network uses participants’ internet bandwidth to help access and collect public web data. Users provide the connection layer, while the web itself supplies the information used to build datasets. As more of the internet becomes difficult to access programmatically, distributed bandwidth becomes part of the infrastructure connecting public web data with AI.

Kaiz 🐂🀄️

13,127 Aufrufe • vor 1 Monat

FRIEDBERG: “With the number of satellites going up with Starlink, and to actually deploy data centers that can output data on those nodes on that network, SpaceX has largely built a backup Internet.”

FRIEDBERG: “With the number of satellites going up with Starlink, and to actually deploy data centers that can output data on those nodes on that network, SpaceX has largely built a backup Internet.”

Chief Nerd

35,203 Aufrufe • vor 7 Monaten

Grass is a decentralized web scraping protocol that rewards users for sharing internet bandwidth — data collected is used to train AI models. On Feb. 14 touch grass hit an ATH in daily data collected – with 1M+ GB scraped 🤯 more on Grass from with Kevin Follonier - MR SHIFT 🦁

Grass is a decentralized web scraping protocol that rewards users for sharing internet bandwidth — data collected is used to train AI models. On Feb. 14 touch grass hit an ATH in daily data collected – with 1M+ GB scraped 🤯 more on Grass from with Kevin Follonier - MR SHIFT 🦁

Solana

26,652 Aufrufe • vor 1 Jahr

Trump: "I'm pretty good at this stuff. If you can go through years of these press conferences and you're a popular president of the United States that won in a landslide, that won all 7 swing states, that got record numbers of votes -- I mean, I guess I'm okay at this stuff."

Trump: "I'm pretty good at this stuff. If you can go through years of these press conferences and you're a popular president of the United States that won in a landslide, that won all 7 swing states, that got record numbers of votes -- I mean, I guess I'm okay at this stuff."

Aaron Rupar

119,940 Aufrufe • vor 7 Monaten

Quick explanation of why you can see something happen a lot, but that doesn't mean it's the only thing happening. Also, I have a lot of new stuff. Link in replies.

Quick explanation of why you can see something happen a lot, but that doesn't mean it's the only thing happening. Also, I have a lot of new stuff. Link in replies.

hoe_math = PsychoMath

28,124 Aufrufe • vor 1 Jahr

Joe Rogan: “I think there's more politically homeless people now than ever.” “They're gonna side with one side or the other based on their opinion about Donald Trump.” “Both teams are filled with scoundrels. You go far enough to the left and far enough to the right, you have the same kind of monster that's just adopted a different ideology.” “Anybody can join the left and anybody can join the right, and there's a lot of mentally ill people out there.”

Joe Rogan: “I think there's more politically homeless people now than ever.” “They're gonna side with one side or the other based on their opinion about Donald Trump.” “Both teams are filled with scoundrels. You go far enough to the left and far enough to the right, you have the same kind of monster that's just adopted a different ideology.” “Anybody can join the left and anybody can join the right, and there's a lot of mentally ill people out there.”

End Tribalism in Politics

25,269 Aufrufe • vor 2 Jahren

Y Combinator co-founder Paul Graham says he's constantly telling founders that all that matters is growth rate, not the absolute numbers. His rationale is straightforward: mathematically, if your growth rate is high enough, it simply doesn't matter what the absolute numbers are. "And the great thing about focusing on growth rate is you can focus on selling to startups for cheap, instead of having to go and do these big deals with big companies that take a long time and make your product stupider."

Y Combinator co-founder Paul Graham says he's constantly telling founders that all that matters is growth rate, not the absolute numbers. His rationale is straightforward: mathematically, if your growth rate is high enough, it simply doesn't matter what the absolute numbers are. "And the great thing about focusing on growth rate is you can focus on selling to startups for cheap, instead of having to go and do these big deals with big companies that take a long time and make your product stupider."

TBPN

82,646 Aufrufe • vor 8 Monaten

jordan peterson explains why being articulate is one of the most valuable skills you can develop. “there is no more exceptional form of the capacity to be dangerous than to be articulate.” “young men are never taught this. why become literate? do you want to be competent and dangerous or do you want to be vague and useless? because those are your options.” “it doesn’t matter what your job is. if you’re a plumber and you’re articulate, you can negotiate with clients, lead your co-workers, advertise your services, and think through your problems.” “our whole culture is based on the supremacy of the word. it’s the word itself that extracts habitable order from chaos and possibility.” “if you want to make yourself into a force to be contended with, there’s no surer path than exceptional literacy.”

jordan peterson explains why being articulate is one of the most valuable skills you can develop. “there is no more exceptional form of the capacity to be dangerous than to be articulate.” “young men are never taught this. why become literate? do you want to be competent and dangerous or do you want to be vague and useless? because those are your options.” “it doesn’t matter what your job is. if you’re a plumber and you’re articulate, you can negotiate with clients, lead your co-workers, advertise your services, and think through your problems.” “our whole culture is based on the supremacy of the word. it’s the word itself that extracts habitable order from chaos and possibility.” “if you want to make yourself into a force to be contended with, there’s no surer path than exceptional literacy.”

gomi

14,985 Aufrufe • vor 3 Monaten

1/ New episode with Andrej Radonjic (Andrej), Co-Founder and CEO of touch grass! We discuss: - Grass as critical infrastructure to power the AI boom - Scaling to 2M+ contributors within months - How Grass is democratizing AI

1/ New episode with Andrej Radonjic (Andrej), Co-Founder and CEO of touch grass! We discuss: - Grass as critical infrastructure to power the AI boom - Scaling to 2M+ contributors within months - How Grass is democratizing AI

Proof of Coverage | Tech’s Creative Agency

298,672 Aufrufe • vor 2 Jahren

Krishna Srinivasan, Founder and CEO of Data Bootstrap and former Google DeepMind researcher, discusses the challenges of data scraping and scalability. With experience at prior Apple, Yahoo, and IBM watsonx, Krishna demonstrates the difference between scraping with a single machine versus leveraging OpenLedger's community nodes. The results speak for themselves, parallelized scraping enables significantly higher efficiency and scale. Data Bootstrap is one of many companies building on our Data Intelligence Layer, and we’re excited to support innovations like these. Watch the video below to learn more.

Krishna Srinivasan, Founder and CEO of Data Bootstrap and former Google DeepMind researcher, discusses the challenges of data scraping and scalability. With experience at prior Apple, Yahoo, and IBM watsonx, Krishna demonstrates the difference between scraping with a single machine versus leveraging OpenLedger's community nodes. The results speak for themselves, parallelized scraping enables significantly higher efficiency and scale. Data Bootstrap is one of many companies building on our Data Intelligence Layer, and we’re excited to support innovations like these. Watch the video below to learn more.

OpenLedger

12,089 Aufrufe • vor 1 Jahr