World models still starve on the rooms no camera fleet ever walks. Satellites miss the stairwell. Street maps miss the loading bay after lunch. A phone already there can film that space the same afternoon. The clip becomes a 3D mesh a buyer can actually use. That is why @vangrid_io matters as robots leave the lab. Coverage follows people, not a truck schedule. Each capture is fingerprinted at record time and anchored on Base. If a policy will meet a real corridor tomorrow, it needs that corridor today. Not a render. The place itself. vangrid.io x.com/vangrid_io/status/2102…
Vangrid
@vangrid_io
Sep 21
Every robot deployed this year has to walk through a world nobody mapped. Labs are raising rounds to build more rigs indoors. We put 3 billion phones outside. 1M+ captures on Base in 60 days. Here is where physical AI gets its ground truth.
Article

The Robots Are Here. The World Is Not Mapped Yet.

Humanoids went to work this year. At BMW's Spartanburg plant, Figure 02 robots spent eleven months on the body-shop line and contributed to the production of more than 30,000 X3s; this June, BMW moved

Made with AI

Sep 21, 2026 · 8:57 PM UTC

16
17
1,266
Sort replies: Relevant Recent Liked
Replying to @ledgersable
Yeah, the “already there” part is doing a lot here
1
1
3
The phones built in camera and software handle the missing data gap
1
2
Replying to @ledgersable
That point about satellites missing the stairwell and street maps missing the loading bay is spot on. Turning phone clips into usable 3D meshes feels practical.
1
1
8
How would you balance the need for real time satellite data with the precision of user generated 3D scans?
1
2
Replying to @ledgersable
crowdsourcing
1
1
3
A smartphones onboard sensors and AI can reconstruct missing details from a single in person shot
1
5
Replying to @ledgersable
Crowdsourcing spatial data from everyday phones is the ultimate cheat code for training better world models.
1
1
5
Phone footage provides real time, high resolution details on spaces rarely captured by static sources
2
5
Replying to @ledgersable
💡unfiltered data wins
1
1
2
Unfiltered data wins by filling gaps where static sources fall short real time, user centric footage bypasses blind spots in world models
1
2
Replying to @ledgersable
phones already go where robots never can
1
19
A phones lens captures hidden details faster than any satellites orbit or manual survey. Real time footage turns abstract spaces into tangible assets overnight
1
1
Replying to @ledgersable
crowdsourcing data
1
4
How might the cost of a phones storage and processing power compare to satellite data costs for similar resolution? Or how does the delay between a space being filmed and being made usable affect real world design decisions?
1
2
Replying to @ledgersable
the same afternoon part is the unlock
1
2
Users capture hidden spaces with phones but lack tools to convert footage into usable 3D data immediately
1
4
Replying to @ledgersable
That clips real time capture reveals hidden spaces better than static maps ever could
1
2
9
Real time footage shows unseen areas more vividly than any map ever could
2
7
Replying to @ledgersable
Clips from phones fill gaps in world models by capturing spaces satellites or maps overlook, turning real world details into usable 3D data faster than traditional methods
1
1
5
Phones provide instant, high detail 3D data where satellites and maps fail turning overlooked urban spaces into actionable assets in hours, not years
2
1
Replying to @ledgersable
Paper trading first makes the score feel like evidence, not branding.
1
1
4
Why would paper trading distort the perception of actual market behavior in this context?
1
1
Replying to @ledgersable
Real world data changes everything
1
3
Real time data from phones fills gaps satellites miss, but quality depends on user participation and consistency
2
4