The value of data always trends to zero. Is data even valuable?
Yes, but only at a certain point. If you’re a big retail e-commerce company, you may want to get data on the prices of your competitors, but if the data is from last week, it’s not really that helpful.
Real time data, is what has value. Unfortunately, accessing this data has historically been difficult. Web2 companies have notoriously used users personal devices for this we’re yet to see a wide scale decentralised solution for the data problem. Well, actually, that’s not true anymore...
Crypto and specifically DePin has brought some very interesting innovations. One of my favourites by far is
@getgrass_io which today has over 2.5m users opting in to collect data from the internet in real time and use it to power AI models. Now I’ve covered Grass many times before so I recommend watching my deep dives or my previous tweets to learn more about how Grass actually works.
My deep dive video on grass -
x.com/0xMughal/status/175846…
Grass is collecting TBs of data. What is the value of it?
Data on the internet comes in many forms, text, images, video etc.
What is interesting is seeing how much this data can cost. The market rate for text is $0.001 per word, whereas with images, companies can be paying $1 and video ranging from a few dollars to even $300 for long form films.
With Grass now collecting enough data in a day to power GPT 3.5, the possible revenue that can be generated here is something to keep an eye on. Whilst we haven’t yet seen which companies are working directly with Grass to purchase this data, we have had confirmation that companies are working with Grass because it’s their only option.
Let’s go into some of the recent upgrades with Grass
Live Context Retrieval - LCR
LLMs are great, they require training and once powered open up so many doors. The one issue however, they don’t have access to real time data.
They need to be constantly trained for this. LCR opens the doors for LLMs to leverage and pull a ton of real time data that is powered by the 2.5m nodes running grass, being able to retrieve anything from the internet in less than half a second. LCR can earn revenue for grass in many different ways too, with companies using it for inference, companies using it for model training and also for ecosystem projects.
Who is even using LCR?
Here’s an example of where protocols can leverage LCR.
x.com/guapmuffin/status/1845…
LCR is what LLMs might need to scale further. One of the best things about it within
@getgrass_io is the fact that its baked into the tokenomics through a market buy model. Client pays in fiat, it gets converted into
$GRASS, if you're bandwidth is used, you get rewarded as a node operator.
Distribution
We all know Grass has a desktop node, but we should soon be able to run Grass on our iPhones. Something
@0xdrej mentioned recently on a podcast was the ‘Grass Box’. A physical hardware device coming soon.
The Grass Token
We're in crypto so it would be prudent of me not to mention the elephant in the room. I honestly, have no idea what the valuation of the grass token will be. For me, I'm more interested in seeing the value capture.
DePin projects have notoriously been great at providing supply and less great at getting demand. Here we have one of the few opportunities where the demand side has utilisation. I know
@0xdrej understands tokenomics well so excited to see what gets cooked in this domain. Give us some good old game theory too!