Observed arrival · 2026-10-26
GlassBench tests the depth-sensing failure robots can’t see
Limit Case says it builds failure data and stress tests for physical AI, beginning with depth-sensing failures on glass and shiny surfaces.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Robotics and depth-perception developers
- Worth noticing
- The homepage reports that one widely used depth camera returned values for nearly every pixel on transparent objects, yet 68.6% were wrong.
Field notes
Limit Case distinguishes a sensor returning no reading from one returning plausible but incorrect geometry, arguing that the latter can leave robot systems unaware of the error. The page says its initial work centers on glass and shiny surfaces, and that failures must be generated physically and calibrated to sensor behavior. It also says common metrics can obscure these silent failures by averaging them with visible ones.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
◎NicheUnusually specific use
One card from the complete issue