Grass

Our mission is to redefine Internet incentive structures

[FalconX @FalconXGlobal]

https://t.co/Gaao9lPd61

[Yan Liberman @YanLiberman]

New Delphi Podcast with Andrej Radonjic @0xdrej, 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

[Grass @grass]

https://t.co/lVn312cCS3

[Grass @grass]

Today, we’re launching our first Live Context Retrieval tool for agents.

Introducing the Grass Contents API — an API that turns any public URL into clean, model-ready HTML, Markdown, or text.

[Grass @grass]

Request access now: https://t.co/CBFBDFbxyT

[Grass @grass]

Grass is designed to keep value in the ecosystem. This is how its organizational structure protects that purpose:

[Grass Foundation @grassfdn]

https://t.co/odd2lvvwbz

[Grass @grass]

Let’s free the internet

[shayon @shayonsengupta]

https://t.co/HvGWr3OsjN

[Grass @grass]

https://t.co/zQVu76ei7C

[Grass @grass]

Revenue generated under customer contracts flows directly to Grass DataCo Ltd., not Wynd Labs. Wynd Labs is a separate vendor: it helps build Grass products and services and engages in sales and marketing in support of Grass DataCo Ltd.'s products and services. This structure exists to ensure that the revenue generated by Grass DataCo Ltd. remains within the Grass ecosystem. We know this has been a point of confusion so we'll be putting out a write up on the topic so people can thoroughly understand the well thought out nature of our structure.

Grass reinvests the majority of revenue. In 2025, Grass bought compute and storage ahead of major cloud price hikes. That single move is cutting network opex by on average $2.5M+/month, and it's a direct reason the business is profitable today. Grass is focused on building a strong token-centric ecosystem.

Financial transparency is part of how Grass builds trust with network participants and token holders.

[Grass @grass]

Grass DataCo Ltd., a wholly owned subsidiary of Grass Foundation, provides data infrastructure to leading AI labs training frontier models.
We're publishing an independent attestation of its revenue through Q2 2026, conducted by Regen Financial.
Full report: https://t.co/yvqFre8YZd

[Grass @grass]

The world is coming to realize what we've always known.
The difference is the data.

[Jasper Lu @lu__jasper]

This is notable. DeepSeek, a lab usually first to pioneer novel algorithms and architectures, is saying that at this point, the ROI of improving data quality far exceeds that of working on novel post-training algorithms.

I think this has already been true for some time for non-lab practitioners. If you're doing llm post-training, 80% of your effort should go into looking at your data.

This means:
- Hiring experts to dig through your RL tasks
- Sifting through rollouts and sft data by hand to remove suspicious samples. Make sure all tasks are actually passable.
- Making sure your data is diverse in both difficulty and category.

[Grass @grass]

GRASS is now live on coinbase․com and in the Coinbase app.

The $GRASS/$USDC spot trading pair is now available.

https://t.co/Gf7DOf9tQw

[Coinbase Markets 🛡️ @CoinbaseMarkets]

Grass (GRASS) is now live on coinbase․com and in the Coinbase app. Coinbase customers can log in to buy, sell, convert, send, receive or store these assets.

Asset availability varies by jurisdiction. https://t.co/nXLFHlmC6s

[Coinbase Markets 🛡️ @CoinbaseMarkets]

Grass (GRASS) is now live on coinbase․com and in the Coinbase app. Coinbase customers can log in to buy, sell, convert, send, receive or store these assets.

Asset availability varies by jurisdiction. https://t.co/nXLFHlmC6s