Pinned Tweet
a robot can learn a task in simulation, but the real world never follows the script perfectly
different streets, changing environments and unexpected obstacles all matter
that’s why real-world data is such an important piece of physical AI
@vangrid_io is building a way to collect location-specific 3D data through smartphones, with captures anchored onchain
the models of tomorrow need better ways to understand the world we actually live in
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
most people already have a feed full of opinions. turning those signals into actual trading decisions is a different story.
@zerufinance and Aspen connect the two by letting users turn selected X and YouTube sources into automated Hyperliquid trades.
through Zaps, qualifying live activity can count toward the leaderboard and its $50K campaign pool. paper trades don't count, and neither do wash trades or instant round-trips.
what interests me is the gap between following a signal and taking a real position. automation makes that gap smaller, but risk doesn't disappear.
the campaign rewards are separate from the planned Zaps airdrop, and eligibility doesn't guarantee a payout.
before connecting a feed, understand the strategy, check the official rules, and know what your automation is allowed to do.
🤝 Paid partnership
most people already have a feed full of opinions. turning those signals into actual trading decisions is a different story.
@zerufinance and Aspen connect the two by letting users turn selected X and YouTube sources into automated Hyperliquid trades.
through Zaps, qualifying live activity can count toward the leaderboard and its $50K campaign pool. paper trades don't count, and neither do wash trades or instant round-trips.
what interests me is the gap between following a signal and taking a real position. automation makes that gap smaller, but risk doesn't disappear.
the campaign rewards are separate from the planned Zaps airdrop, and eligibility doesn't guarantee a payout.
before connecting a feed, understand the strategy, check the official rules, and know what your automation is allowed to do.
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
a robot can learn a task in simulation, but the real world never follows the script perfectly
different streets, changing environments and unexpected obstacles all matter
that’s why real-world data is such an important piece of physical AI
@vangrid_io is building a way to collect location-specific 3D data through smartphones, with captures anchored onchain
the models of tomorrow need better ways to understand the world we actually live in
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
a leaderboard can show who traded the most. the harder part is knowing which activity actually counts.
@zerufinance uses Zaps to track qualifying activity in the Aspen campaign, with verified volume affecting your rank and potential share of the $50K reward pool.
i like the distinction between your existing Zaps balance and campaign performance. your past wallet activity gives you a starting point, while new qualifying trades determine how you compete.
wash trading and sybil activity don't count, and a high rank alone doesn't guarantee a payout.
the useful thing to check first is your starting balance and the official reward rules. knowing how the scoring works matters before chasing a number.
🤝 Paid partnership
a leaderboard can show who traded the most. the harder part is knowing which activity actually counts.
@zerufinance uses Zaps to track qualifying activity in the Aspen campaign, with verified volume affecting your rank and potential share of the $50K reward pool.
i like the distinction between your existing Zaps balance and campaign performance. your past wallet activity gives you a starting point, while new qualifying trades determine how you compete.
wash trading and sybil activity don't count, and a high rank alone doesn't guarantee a payout.
the useful thing to check first is your starting balance and the official reward rules. knowing how the scoring works matters before chasing a number.
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
gm gVangrid CT
one thing i like about @vangrid_io is that the data collection starts with devices people already carry
instead of relying only on expensive specialized hardware, smartphones can help capture real environments when specific data is needed
that creates a practical path for building the 3d datasets physical ai will need to understand the world beyond a screen
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
hi guys, kinda noticed the blur happens way earlier than i thought when looking at how @vangrid_io capture works.
you open the app in browser, allow camera, and just start recording while walking around a place slowly...
as the video is going, the phone finds faces and car plates and blurs them on the spot.
phone blurs faces on device before the video is encoded.
clear version never leaves the phone, only the blurred clip gets uploaded later...
tried a quick one and the blur popped up on faces in the view almost right away, not after the upload or anything.
then when you stop recording it does this quality check and rejects if too short or unreadable, tells you why.
your original stays with you, they only hand over the finished model.
does that quality check show the reason immediately after you hit stop, or is there a step before the clip is ready for a @vangrid_io Bounty?
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
Noticed something interesting while testing how @vangrid_io capture works i expected the privacy part to happen after recording.
But the phone seems to handle that much earlier.
You start recording in the browser walk around and let the camera pick up the surroundings.
Faces and license plates get blurred directly on the device before the video is encoded.
That detail matters more than i first thought.
Because the original footage stays on your phone while only the processed clip gets uploaded.
I tried to picture what would happen if that blur step ran on a server instead.
The raw footage would already have left the device before the privacy protection kicked in.
Here the order is different.
The capture gets checked after you stop and clips that are too short or unreadable can be rejected with a reason.
So it is not just about collecting useful spatial data it is also about deciding what should never leave the phone in the first place.
Still curious about one thing though.
When a clip fails that quality check can you fix the same recording or do you have to capture everything again for the @vangrid_io bounty?
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
a mapping van has a lead time. a phone already in the pocket does not. that is the supply chain @vangrid_io never had to build.
most sensor networks start with a box. you design it, wait for the factory, ship it, wait for someone to install it, then hope the density shows up. every step is inventory and delay. the phone already walks the street every day. open the app, walk the named place, blur faces and plates on-device, and the geometry comes back with a fingerprint at record time.
no unit to manufacture. no pallet to track. no installer to schedule. the node is the device the contributor already carries.
if the place needs ground truth this week, the supply chain is already there.
@vangrid_io
gm gVangrid CT
one thing i like about @vangrid_io is that the data collection starts with devices people already carry
instead of relying only on expensive specialized hardware, smartphones can help capture real environments when specific data is needed
that creates a practical path for building the 3d datasets physical ai will need to understand the world beyond a screen
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
good morning everyone
a new day in web3, another chance to build something meaningful.
@NucleusCodes makes me think about how much value exists beyond wallet size. Consistent contributions, useful knowledge, and real community involvement deserve recognition too.
your reputation is more than a number. keep learning, keep contributing, and let your work speak.
have a productive day everyone 🤍
good morning everyone
a new day in web3, another chance to build something meaningful.
@NucleusCodes makes me think about how much value exists beyond wallet size. Consistent contributions, useful knowledge, and real community involvement deserve recognition too.
your reputation is more than a number. keep learning, keep contributing, and let your work speak.
have a productive day everyone 🤍
𝗦𝘆𝗶𝗳 retweeted
your trading history and your next trade are two different things.
@zerufinance connects them through zScore and Zaps. past wallet activity can give you a starting Zaps balance, while new qualifying trades can add to it.
the Aspen campaign adds a separate competition on top, with $50K in rewards tied to verified Hyperliquid volume.
what matters is the distinction: existing history helps establish your starting point, but campaign rewards depend on the published rules and qualifying activity.
reputation shouldn't need to start from zero every time you try something new.
now the interesting part is seeing how well this model separates genuine trading from volume created just to climb a leaderboard.
🤝 Paid partnership
your trading history and your next trade are two different things.
@zerufinance connects them through zScore and Zaps. past wallet activity can give you a starting Zaps balance, while new qualifying trades can add to it.
the Aspen campaign adds a separate competition on top, with $50K in rewards tied to verified Hyperliquid volume.
what matters is the distinction: existing history helps establish your starting point, but campaign rewards depend on the published rules and qualifying activity.
reputation shouldn't need to start from zero every time you try something new.
now the interesting part is seeing how well this model separates genuine trading from volume created just to climb a leaderboard.
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
GM gVangrid CT
Data Is The Boring Part Of AI Until It Runs Out. Then It's The Only Part That Matters
language models got the internet. robots don't get a free version of a loading bay or a stairwell, someone has to go capture it
@vangrid_io is building that layer. phones capture real places, captures are anchored onchain on Base, and USDC settles when work is accepted
1,283,975 captures and 629,299 USDC settled on the explorer right now. live since July, so you can check it instead of trusting a roadmap
explorer.vangrid.io
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
gvangrid ☕
Can Sit Underneath the Builds : Vangrid
@vangrid_io is the tech that can sit underneath:
Different teams are going to build the robots <•>
Different teams will build the agents <•>
Different teams will build the models <•>
But they all will run into the same problem
What is happening in the physical world right now.
That is where a system like Vangrid can sit underneath the stack.
Not competing to build the robots or the agents.
Just becoming the layer they can call when they need fresh ground data.
With a simple system :>
→ If the data exists, use it <•>
→ If it is missing, create it <•>
A lot of builds could end up needing the same door into reality.
Vangrid sets itself up as that door <•
Time to find the next missing data set <•
Dive in <•>
𝗦𝘆𝗶𝗳 retweeted
was wondering how a phone video of some random alley is meant to hold up later as actual ground truth...
with @vangrid_io the capture runs right in the browser.
you walk the spot slowly, the phone detects faces and plates while recording and blurs them on the device before the video gets encoded.
original frames never leave the phone.
looked at the guidance and it says keep it steady, hit every side you can, skip the fast pans or you get holes in the 3D model...
phone blurs faces on device then the clip gets fingerprinted for the chain.
those fingerprints get batched and the root lands on Base about every 15 minutes.
@vangrid_io has the Explorer where you can see the latest trees and their capture counts.
when the clip stops uploading, does the quality check message appear right there or do you have to refresh something?
GM gVangrid CT
Data Is The Boring Part Of AI Until It Runs Out. Then It's The Only Part That Matters
language models got the internet. robots don't get a free version of a loading bay or a stairwell, someone has to go capture it
@vangrid_io is building that layer. phones capture real places, captures are anchored onchain on Base, and USDC settles when work is accepted
1,283,975 captures and 629,299 USDC settled on the explorer right now. live since July, so you can check it instead of trusting a roadmap
explorer.vangrid.io
🤝 Paid partnership
𝗦𝘆𝗶𝗳 retweeted
A robot following a perfect route still runs into equipment that wasn't there yesterday. That's not a routing problem. That's a data problem, and most of the industry is solving the wrong one.
Everyone's racing to build smarter navigation models. Fewer people are asking where the model's actually getting its picture of the world, and how stale that picture already is by the time the robot shows up.
Vangrid's approach skips the specialized mapping hardware entirely. Someone with a phone walks through a loading bay, a warehouse aisle, a street corner that changed since the last scan. That footage gets filtered for privacy on-device, then stamped with a cryptographic fingerprint that ties back to an onchain record, so the capture isn't just trusted, it's checkable.
The bounty system is what makes this scale past a hobby project. Organizations that actually need current ground-level data on a specific location can fund a capture request directly. Contributors get paid in USDC when the submission clears the bar. Nobody's capturing blind and hoping someone finds it useful later.
That matters more for autonomous logistics than another model upgrade would. A warehouse robot doesn't need better reasoning about a stale map. It needs a map that's actually still true.
Smarter models get all the attention. Current, verifiable ground truth is the part nobody's shipping at scale, and it's the part that actually decides whether the robot gets stuck or not.
@vangrid_io
🤝 Paid partnership
