The return of internal criticism of the Tesla method
One of the figures who helped build Tesla's first Autopilot system is now questioning the technical foundations of the strategy adopted by the company. Sterling Anderson, who headed the Autopilot project in 2015 and 2016 and previously led the Model X program, described the abandonment of radar in favor of cameras alone as a form of obstinacy that disregards the laws of physics. According to the source's account, this is the most direct attack he has so far directed at so-called pure vision.
The stance comes at a time when Tesla is deploying the Cybercab, a vehicle based exclusively on cameras, in a driverless taxi service in Austin. Anderson spoke on October 27 in an interview with The Verge, just over a week after a technology presentation organized by General Motors in New York.
The argument: some situations cannot be handled without radar
Asked whether General Motors might eventually follow the path traced by Tesla, Anderson answered bluntly: returning to vision alone, he argued, from the standpoint of physical principles would inevitably lead to problems under certain conditions. His reasoning rests on two limitations he attributes to cameras: they cannot penetrate thick fog and they do not measure instantaneous speed as radar does.
To this he adds an economic consideration: radar is inexpensive, and doing without it would have no justification. Anderson also argues that training lidar and cameras together could accelerate the improvement of vision systems. The argument directly targets Tesla's position that cameras and data collected by its fleet are sufficient for autonomous driving.
Who is speaking and what line he is defending
The source stresses that Anderson is not speaking in the interest of a lidar startup seeking visibility. After his time at Tesla, he co-founded Aurora, a company active in autonomous driving, and in 2025 he was hired by General Motors as chief product officer. Bloomberg reported the previous year that Mary Barra considered him among the most likely candidates to succeed her as chief executive officer.
His approach at General Motors is avowedly opposed to Tesla's: the Cadillac Escalade IQ is scheduled for 2028 with a driver-assistance system combining lidar, radar and cameras. In the interview Anderson also distanced himself from promotional phrases such as eyes off, noting that a very effective system leads those at the wheel to relax their attention and trust it more than they should.
The 2021 decision and what remains open
The historical reference is May 2021, when Tesla removed radar from Model 3 and Model Y, adopting its own vision-based solution. According to former employees, engineers at the time tried to dissuade Elon Musk, who confirmed the choice. Musk had told Electrek that a high-resolution radar would have been preferable to vision alone, but that such a product did not yet exist.
The material merely notes that several high-resolution radars are now available on the market, without specifying which ones and without independent evaluations of the Cybercab's performance. It therefore remains to be verified in the field how much the limitations described by Anderson affect real operations and whether General Motors will maintain the plan announced for 2028.
