📬 Remix Reality Insider: Inside the Sensing Stack

Your weekly briefing on the systems, machines, and forces reshaping reality.

🛰️ The Signal

This week’s defining shift.

Sensing is being designed as a complete system. The center of gravity has shifted from tuning individual cameras or lidar units to making sure the whole stack works together in real conditions. You can see it in the hardware choices and how these products are being packaged and sold.

This week’s news surfaced signals like these:

• Waymo introduced its 6th-generation Driver with a redesigned sensing suite that balances cameras, lidar, radar, and audio around cost, weather performance, and multi-vehicle deployment.
• Ouster acquired StereoLabs, bringing stereo vision hardware and perception software into its lidar business and repositioning itself around an integrated sensing and perception platform.

Why this matters: Where and how sensing systems run is now shaping how these stacks are designed.

🧠 Reality Decoded

Your premium deep dive.

Our latest editorial from Richard Pallardy looks at camera traps and the use of computer vision as an example of real-world perception systems. It shows how AI can turn chaotic, low-quality image data into something useful, and why the real breakthrough is in filtering noise, not building better cameras.

Three takeaways from this piece:

• Filtering is as important as understanding: Most images get thrown away so people and models can focus on what actually matters. Deciding what to ignore is as important as recognizing what’s there.
• Real systems are trained on failure cases, not clean examples: Camera trap images are inconsistent and full of edge cases. Models only hold up when they are trained on large volumes of imperfect, real-world data.
• Edge cases set the ceiling on performance: Rare species and individuals are still difficult to identify, especially with limited variation. Even at scale, systems default to the common case and miss the rare ones.

Key Takeaway:
Working perception is less about perfect models and more about surviving messy reality. In practice, systems live or die on reliability, not on how clever they are.

📡 Weekly Radar

Your weekly scan across the spatial computing stack.

🚕 Baidu and Uber Expand Autonomous Ride-Hailing to Dubai

• Apollo Go autonomous vehicles will launch on the Uber app in select areas of Dubai’s Jumeirah district in the coming month.
• Why this matters: Dubai marks a clear step in Apollo Go’s push beyond its home market. With London plans already announced and a dedicated operations hub now open in the city, Baidu is positioning Apollo Go as an international operator.

🦾 Vention Launches Generalized Physical AI Pipeline for Factory Robots

• Vention introduced GRIIP, an end-to-end physical AI pipeline designed to deploy autonomous robot cells in unstructured manufacturing environments.
• Why this matters: Less custom engineering, more out-of-the-box autonomy. If GRIIP delivers, it shifts robotics from carefully programmed tasks to production cells that can adapt and keep running.

👓 EssilorLuxottica Reports 7 Million AI Glasses Sold in 2025

• EssilorLuxottica sold more than 7 million AI-glasses units in 2025, with all regions and brands contributing to the total.
• Why this matters: When the world’s largest eyewear company starts tying its growth directly to wearables, it is no longer a side project, it’s a category shift.

🔍 Yokogawa and ANYbotics Join Forces on Integrated Robot Inspection for Industry

• ANYbotics’ autonomous inspection robots will now be managed directly through Yokogawa’s OpreX plant control platform.
• Why this matters: This partnership makes it possible to manage autonomous inspection robots through the same systems used to run industrial plants. It’s a practical move toward automating routine inspection in hazardous environments.

🌏 NVIDIA and Samsung Back Odyssey’s Push Toward a General-Purpose World Simulator

• Odyssey received funding from NVentures and Samsung Next to expand development of its world modeling systems.
• Why this matters: Backing from NVentures and Samsung Next signals that major platform players see world models as reaching a critical inflection point.

⛏️ Propeller Acquires Spacesium to Automate Spatial Mapping for Mining Sites

• Propeller has acquired Spacesium to embed geospatial algorithms into its 3D mapping platform.
• Why this matters: With this acquisition, Propeller is moving from capturing site data to using geospatial AI to automatically analyze it, starting with haul road safety in mining.

📷 Ouster Acquires StereoLabs to Add Vision to Its Lidar Business

• Ouster completed the acquisition of StereoLabs on February 4, 2026, bringing stereo camera hardware and perception software into its product portfolio.
• Why this matters: Bringing vision and lidar together matters because real-world systems increasingly rely on both to perceive depth, movement, and context. Combining mature camera hardware, perception software, and lidar under one platform reflects how sensing stacks are actually being built and deployed today.

🚗 Waymo Introduces 6th-Generation Driver as Foundation for Scaled Expansion

• Waymo is launching its 6th-generation Driver into fully autonomous service.
• Why this matters: This Driver is designed to expand. It reflects real-world experience and is built to be deployed at scale.

🌀 Tom's Take

Unfiltered POV from the editor-in-chief.

Generative AI is powerful, but large language models are out of step with reality. They work on copies of the past, depending on what we have documented and stored online. That means there is always a gap between how they see the world and what is actually happening in it, shaped by incomplete data and by the biases of the people who created that data.

The gap shows up in what these systems miss, in what they get wrong, and in what they confidently make up. Even as the models improve, they are still looking backward and trying to reconstruct the world from records instead of seeing it as it changes.

That is why physical AI matters. When you add sensors, AI moves off the screen and into the world. Cameras, lidar, microphones, and other sensors let these systems collect information in real time, not just from what was recorded before. With the right access to that data, AI can build a more complete and useful picture of what is going on. It gets the gift of context.

That is why sensing matters. It is what turns AI from something that explains the world into something that can actually understand and operate in it.

🔮 What’s Next

3 signals pointing to what’s coming next.

1. Wearable robotics as standard work equipment
   Exoskeletons are being designed to sit alongside other tools and protective gear in manufacturing, logistics, and construction. Comau’s MATE-XT GO is light, quick to put on, and certified for regulated environments, which makes it fit naturally into daily industrial workflows.
2. Spatial mapping as a part of how sites are run
   Spatial data is guiding day-to-day decisions in the field. Propeller is turning site maps into a system that automatically checks safety and compliance using geospatial AI.
3. Humanoid robots as an industrial production category
   Apptronik is putting nearly a billion dollars behind Apollo to expand manufacturing, training, and deployment with partners in logistics and manufacturing. The focus is on factories, supply chains, and repeatable rollout, not one-off pilots.

Know someone who should be following the signal? Send them to remixreality.com to sign up for our free weekly newsletter.

📬 Make sure you never miss an issue! If you’re using Gmail, drag this email into your Primary tab so Remix Reality doesn’t get lost in Promotions. On mobile, tap the three dots and hit “Move to > Primary.” That’s it!

🛠️ This newsletter uses a human-led, AI-assisted workflow, with all final decisions made by editors.