$GRASS : Review 📜
What if the 95% of your home internet connection you never use could be rented out to the AI labs that can no longer crawl the web without getting blocked?
Meet Grass, a decentralized data network that turns idle residential bandwidth into an AI data pipeline. Millions of ordinary users run a browser extension, and their connections become the eyes that let AI labs read the open web from real homes rather than blacklisted data centers. It is one of the only DePIN networks that appears to be generating serious revenue.
Let's explore where the revenue is real and where the supply overhang
bites. 👇
⚪ Grass at a Glance
Marketplace Insight: Grass sits at the crossing point of DePIN and the AI data bottleneck, and it may be the rare DePIN with real commercial pull. A July 2026 holder call reported roughly $52M in second-half revenue, an annualized run rate near $104M, which would put it among the highest-earning DePIN networks in existence. The necessary qualifier travels with that number every time: it is self-reported by the project, not independently audited, and no third party has verified it. The clean contrast is that Grass's structural advantage, residential bandwidth contributed at near-zero marginal cost, is also its structural liability, since the same design puts the network's economics in the hands of users who can uninstall it in an afternoon.
⚪ Mission
Grass exists to break the data bottleneck in AI. The argument is that the open web is being closed off, since AI labs crawling from data-center IPs get rate-limited, geoblocked, or fed stale caches, while the companies that can still scrape at scale do it by quietly embedding SDKs in free apps without telling users. Grass's answer is to make the arrangement explicit and paid: users knowingly contribute bandwidth, receive rewards for it, and the data carries a cryptographic proof of where it came from.
🔵 A Brief History
Grass began as a reaction to something its founder saw and disliked. Andrej Radonjic had been running a web-scraping infrastructure business when customers began asking for residential IPs, and he learned that large companies were harvesting real people's bandwidth by sneaking SDKs into free applications. He left a PhD in computational and applied mathematics to build the consensual version of that model, founding Wynd Network in Toronto with two co-founders and starting to build in early 2023.
The network grew fast on a simple pitch: install an extension, earn points. By the 2024 mainnet period it claimed over two million users, funded through a $3.5M seed and a Series A led by Polychain and Hack VC. The token generated in October 2024 with one of the largest airdrops in Solana's history, reaching roughly 2.2 million wallets, and the token peaked near $3.90 within days before beginning a long decline.
Through 2025 and 2026, the project pushed to convert a points economy into a real business. It rebuilt itself as a sovereign data rollup with a ZK processor for data provenance, shipped Live Context Retrieval for real-time AI queries, and raised a $10M bridge round co-led by Polychain and Tribe. A July 2026 holder call reported roughly $52M in H2 revenue. The same period brought a genuine community rupture, with users reacting badly to a team podcast appearance and reports of app uninstalls and token selling, a reminder that in this model the supply side is also the shareholder base.
🔵 Ecosystem Narrative
The organizing idea is that if AI needs the open web, someone has to own the pipes that reach it, and that someone can be the users themselves.
➛ Residential node network. 2.5M-plus nodes across 190 countries, run through a browser extension or app, giving AI labs a geographically diverse view of the web that no single data center can replicate.
➛ Sovereign data rollup. Nodes fetch, routers relay, validators verify, and a ZK processor generates proofs that the scraped content was not tampered with, with provenance recorded on Solana.
➛ Live Context Retrieval. Real-time data retrieval for AI models that need current information rather than a stale training set, which the founder has called where most of the network's value accrues.
➛ Data provenance as a moat. Cryptographic proof of where training data came from, which becomes more valuable if AI-data transparency rules harden under frameworks like the EU AI Act.
➛ Revenue through the Foundation. Customer payments flow to the Grass Foundation, which the team has described as directing value back to the token, the intended link between commercial traction and holders.
➛ Cost asymmetry. Bandwidth contributed by users carries near-zero marginal cost, in contrast to centralized scrapers like Bright Data that must buy commercial IP ranges and server capacity.
⚪ Token Utilities
$GRASS is a utility token whose demand is meant to track data purchases rather than governance.
➛ Node rewards: operators earn the token for contributing bandwidth, weighted by uptime, geography, and data quality.
➛ Data payments: buyers pay for retrieval and datasets, with fiat payments routed into the token for network operations.
➛ Staking: holders stake to validators to help secure the network and share in protocol rewards.
➛ Governance: holders vote on network parameters, with a decentralized validator committee and slashing on the roadmap rather than fully live.
⚪ Key Features
➛ Residential bandwidth sharing through a lightweight browser extension or app, the lowest barrier to entry in DePIN.
➛ 2.5M-plus nodes across 190 countries, delivering geographic diversity that data-center scrapers cannot match.
➛ ZK-verified data provenance, giving buyers cryptographic proof the data was not manipulated.
➛ Live Context Retrieval for real-time AI queries rather than static training sets.
➛ Built on Solana as a sovereign data rollup, keeping raw data off-chain and proofs on-chain.
➛ Fixed 1B supply with no perpetual inflation.
🔵 Meet the Team
Grass runs a two-part structure worth spelling out. Wynd Labs, the Toronto development company, builds the network, while the Grass Foundation is the contractual entity that signs customers and holds the revenue, which is where the token's value-accrual claim actually sits. The founding team was pseudonymous early on, and Radonjic has spoken publicly about choosing to dox himself.
▶️ Core Members:
➛ Andrej Radonjic [ @0xdrej ] - Co-Founder and CEO, Wynd Labs | Left a PhD in computational and applied mathematics at York University, with an engineering physics degree from McMaster, after building a web-scraping infrastructure business. Founded Grass explicitly as a consensual alternative to companies harvesting user bandwidth through hidden SDKs. Leads strategy, the Live Context Retrieval thesis, and the commercial motion.
➛ Grass Foundation | The load-bearing non-person entity. It is the contractual party that customers pay, and the team has stated that all customer revenue flows to it and that the token is the primary vehicle for value accrual, which makes the Foundation the hinge between the business and the holder.
➛ Wynd Labs | The Toronto development company behind the network, backed by Polychain across three consecutive rounds, plus Tribe Capital, Hack VC, and Delphi Digital.
🔵 Ratings
➛ Use Case: ★★★★ (4/5). Grass targets a genuine bottleneck, AI labs increasingly cannot crawl the open web from data-center IPs, and it has built real scale against it: millions of residential nodes across 190 countries, thousands of terabytes delivered, ZK-verified provenance, and a reported revenue run rate that would rank it among the highest-earning DePIN networks. The cost asymmetry against centralized scrapers is structurally real. The 1-point deduction is durability and dependence. The moat is a user base that can uninstall the extension at will, as the 2026 backlash demonstrated, the buyer list is concentrated and undisclosed, and the entire model rests on the legal and ethical footing of residential scraping, which is a live regulatory question rather than a settled one.
➛ Tokenomics: ★★★ (3.5/5). The fixed 1B supply with no perpetual inflation is a real strength, and if customer revenue genuinely routes through the Foundation to the token, the value-accrual story is better than most DePIN peers. The 2-point deduction is distribution and overhang. Insiders hold the majority, with 25.2% to early investors and 22% to contributors, cliff vesting runs into 2028, and cliff unlocks release in single large blocks rather than smoothly. The token is down roughly 86% to 90% from its high, and the revenue underpinning the bull case is self-reported rather than audited, so the strongest argument for the token rests on a number no third party has checked.
➛ Audits: ★★★✦ (3.5/5). This is the honest ceiling on what is disclosed. I could not verify a named third-party smart-contract audit or a CertiK Skynet security score for Grass, and without a named firm the score caps at 3.5 by a consistent standard. The mitigants are real but partial: the token is a straightforward fixed-supply Solana asset, the ZK processor is designed specifically to make data integrity verifiable, and the network has operated at scale since 2024 without a headline exploit. A disclosed audit from a named firm, or a verified Skynet score, would move this to 4 immediately.
➛ Community: ★★★✦ (3.5/5). Grass has extraordinary reach for a DePIN, with millions of node operators, one of Solana's largest airdrop distributions, and a genuinely global footprint. The 1.5-point deduction is that reach is not the same as loyalty, and this community proved it. The user base was assembled through points farming and airdrop expectation, and the mid-2026 backlash saw users uninstalling the app and selling into the market over communications from the team, which is a structural vulnerability specific to this model, since in Grass the community is also the infrastructure.
🔵 Conclusion
Grass has built something few DePIN projects manage: a network with genuine scale and, by its own reporting, genuine revenue. Millions of residential nodes, real geographic coverage that centralized scrapers cannot buy, ZK-verified data provenance, and a product that targets one of the sharpest bottlenecks in AI, the fact that the open web is closing to the labs that need it most.
The risks are real and worth naming. The revenue figure that anchors the entire bull case is self-reported and unaudited. Insiders hold the majority of a supply that keeps unlocking in cliffs through 2028. No named security audit is disclosed. The token trades close to 90% below its high. And the network's infrastructure is a user base that can, and recently did, walk away over a communications misstep.
But the bull case is clean. If AI models need fresh data from a web that increasingly refuses to be crawled, and if provenance becomes a compliance requirement rather than a nice-to-have, then a network of millions of consenting residential nodes with cryptographic proof of sourcing is holding something the labs cannot easily build themselves. Grass has the scale and appears to have the customers.
What it still has to prove is that the numbers survive outside inspection, and that the people who are the network stay when the points run out.
