Pod Notes #2: Grass Network

Week 2 of Pod Notes, episode 55 of Exit Liquidity. This week, we chatted with Drej from Grass about the crypto/AI stack, monetizing idle bandwidth, and how Grass grew to be the largest crypto product by users on the market right now.

• Drej got his start in tradfi, was naturally drawn to on-chain derivatives early on

• Exit Liquidity Story - participated in a DOGE faucet in 2014, lost a few million dollars because he couldn’t access his university email after he left uni. Tried to get it back but then found out the wallet got hacked anyways

• Wynd is the team behind Grass, Grass is an extension that sells users unused bandwidth to AI corporations, generally to be used for training LLMs (gud read here)

• Machine learning models and LLMs are super data-hungry, they need structured data and a buttload of it - Grass provides it to them

• Socrates - New product, not much info yet but will start to expand Grass’ role in the AI stack beyond just data collection

  • “If you're like the only one in your space that actually has the tools to get that oil, why not capture the entire value of it? If there's no one else out there that's capable of getting or drilling for oil the way you're able to, you may as well go and refine it, you may as well go and sell it, you may as well do these other things with it.”

• How did Grass get so big?

  • Unique to Grass - zero upfront capital → points go up. You already pay for your internet/device, Grass uses a very very tiny amount of those resources

  • Referral program allowed them not to have to spend much on influencers, they just created a referral program that heavily incentivizes referring

  • There’s a good number of non-crypto people on Grass - this will be their crypto onboarding

  • Extensions are set-and-forget

• Privacy: Grass does not give a shiet about your data, the extension cannot see anything. It’s like setting up an antenna on the roof of your house, it’s not actually pulling anything from inside. All it’s doing is recording info from public websites

• Biggest challenge: continuing to scale

• Solana Saga phone - was the perfect testing ground for their mobile app, can easily roll back features if needed. Already on 2/3rds of all Saga phones

• Android launch coming soon

• LLMs can be fine-tuned to very specific tasks and they’ll do a very good job, the issue is that for larger tasks you need to take the output of one of those and feed it into another. Creates the possibility of a game of telephone where the odds of getting a bad input is much more likely. It makes more sense to have a more focused approach for LLMs/Agents as of right now for that reason

• They could be interested in exploring compute opportunities as it’s not very resource-intensive, but only when it becomes additive to their network (only focused on data collection rn)

• There are tons of teams building in AI/crypto that deserve credit - he likes ritual specifically

• “The fact that we launched on the Saga phone is a very strong hint towards where we're launching this project.”