The journey toward fully autonomous vehicles is fraught with challenges, a reality underscored by a recent incident at Rivian's Palo Alto office. During the company's "Autonomy & AI Day," a robot navigating the cafeteria unexpectedly halted, displaying "I'm stuck" on its screen. While not a Rivian product, the struggling droid offered a poignant, if unintended, prelude to the day's overarching message: achieving true self-driving capability is incredibly difficult.
This sentiment resonated hours later during a 15-minute demonstration of Rivian's new self-described "Large Driving Model" (LDM) in a 2025 R1S SUV. The electric vehicle, equipped with Rivian's automated driving software, navigated a winding route near the company's campus. The system showcased its capabilities by stopping at lights, handling turns, and slowing for speed bumps without relying on pre-programmed rules. However, the demo was not without its moments of human intervention.
During the drive, the R1S notably braked hard when a Tesla Model S ahead slowed to turn into a rival company's lot, requiring a near-intervention from the Rivian employee in the driver's seat. There was also one actual disengagement when the employee took control through a single-lane section of road due to tree-trimming. While these were minor incidents, they weren't isolated, with other demo rides reportedly experiencing similar disengagements. Despite these hiccups, the software performed commendably for a system not yet ready for commercial release, especially considering Rivian's recent strategic overhaul.
A Quiet Pivot to AI in 2021
Rivian's current autonomous driving efforts represent a significant shift from its previous approach. CEO RJ Scaringe revealed that the company's old system was "very deterministic" and "structured," with every vehicle action resulting from "a prescribed control strategy written by humans."
However, witnessing the rapid advancements in transformer-based artificial intelligence in 2021, Scaringe made a quiet but decisive move. He "reconstituted the team and started with a clean sheet," redesigning Rivian's self-driving platform for an "AI-centric world." After extensive development, this new, ground-up driving software was launched in 2024 on Rivian's second-generation R1 vehicles, which utilize Nvidia's Orin processors. Scaringe noted that dramatic progress only began "once the data started really pouring in."
Rivian is now betting on its Large Driving Model (LDM), which is trained on vast amounts of fleet data, to rapidly accelerate its autonomy roadmap. The company aims to roll out "Universal Hands-Free" driving in early 2026, allowing Rivian owners to remove their hands from the wheel on 3.5 million miles of roads across the U.S. and Canada, provided visible lane lines are present. By the latter half of 2026, Rivian plans to enable "point-to-point" driving, a consumer version of the demo experienced by the reporter.
The 'Eyes Off' to 'Hands Off' Challenge
A significant milestone for Rivian will be the launch of its smaller, more affordable R2 SUVs by the end of 2026. These vehicles will ditch Nvidia chips in favor of a new custom autonomy computer, unveiled at the event, along with a lidar sensor. This advanced hardware is designed to eventually allow drivers to take both their hands and eyes off the road, moving closer to true Level 4 autonomy where the driver is not expected to retake control.
However, this rollout presents a near-term challenge. The custom autonomy computer and lidar won't be ready until months after the R2 goes on sale. This means early R2 customers who desire "eyes-off" driving capabilities will need to wait. Given the R2's critical importance for Rivian, especially following declining sales of its first-generation vehicles, the company needs strong initial sales.
Scaringe addressed this timing mismatch directly, stating, "When tech is moving as fast as it is, there's always going to be some level of obsolescence." He emphasized transparency, assuring that early R2s will still receive Rivian's promised "point-to-point" driving, which will be hands-off but not eyes-off. He acknowledged that some customers might choose to wait for the more advanced hardware, while others will opt for the R2 immediately, potentially upgrading later. "Fortunately, there's so much demand backlog for R2 that we think, by being upfront with this, customers can make the decision themselves," he added, noting that "the timeline of the vehicle and the timeline of the autonomy platform are just not perfectly aligned."
Scaringe's long-term vision for Rivian's autonomous capabilities remains ambitious. In a 2018 interview, he shared a goal of vehicles being so capable that they could meet a driver at the end of a hiking trail, a "pie-in-the-sky promise" that resonated with Rivian's adventure-oriented brand. Scaringe still believes this use case is possible within the next few years, particularly as the company tests and builds its more capable R2 vehicles and trains its LDM on trickier roads without guiding features like lane lines.
While the focus isn't currently on it, Scaringe believes that once Level 4 autonomy is closer, expanding the operational design domain (ODD) to include "dirt roads, off road" will be "easy." However, he clarified that Rivian is "not putting any resources into rock crawling autonomously," such as navigating Hell's Gate in Moab. "But in terms of getting to the trail head? For sure," he confirmed, highlighting a practical, adventure-focused application of their future autonomous technology.
