Meta’s OpenAI raid reminds me of the self-driving arms race
We’ve seen what happens when tech giants fight over researchers. Hint: it didn’t end quietly.
Meta has been in the news this week for poaching researchers across the industry, including from OpenAI.
The news comes on the heels of a major announcement from Zuckerberg. On Monday, the Meta CEO sent a memo to staff introducing the company’s new superintelligence team, which will be helmed by Alexandr Wang, formerly of Scale AI, and Nat Friedman, who previously led GitHub. The list of new hires also included a number of people from OpenAI, including Shengjia Zhao, Shuchao Bi, Jiahui Yu, and Hongyu Ren. OpenAI’s chief research officer, Mark Chen, told staff that it felt like “someone has broken into our home and stolen something.”
On his brother’s podcast, Sam Altman says that the offers have approached $100 million per year, a number also reported by Wired.
Every single article on this has the same 9 facts, and I certainly don’t have anything to add. So let’s go in a different direction: what happened the last time that tech talent poaching was in the news? How did it end up?
The biggest talent poaching spree in recent memory was in the self-driving car industry in the late 2010s1.
Back then, everyone was convinced that self-driving cars were close to being a solved problem, and they were spending as much money as possible to be the industry winner by the beginning of the 2020s.
Google had a self-driving program called Chauffeur. The engineers on Chauffeur believed that their work could be worth billions, so at one point they rebelled. They threatened to leave the company together and get outside venture funding to build their own efforts and reap the rewards themselves.
So Page authorized the creation of a new kind of compensation at Google. The idea was to motivate the team with the sort of incentives they would enjoy at their own startup. Chauffeur's members would remain employees, with regular salaries. The real money came in their bonuses. Every four years, Google would determine the value of the project and pay each teammate a given percentage of the sum, based on their role. Anyone who left before the four-year mark would get nothing.
The problem, of course, is that the valuation for Chauffeur/Waymo skyrocketed. So they got huge payouts (one engineer named Anthony Levandowski2 reportedly got $120 million) and in 2016, with the next payout 3 years into the future, everyone had enough “fuck you” money that there was no reason to wait for the next payout. Many left for more immediate paydays. I heard rumors at the time that some people were just taking their $20 million payouts and retiring, but I can’t find any official sources that confirm that idea. So treat that like the unsubstantiated rumor it is!
Many of the engineers left. For example, Anthony Levandowski left and founded a commercial trucking company called Otto, which was acquired by Uber that same year. A few years later, Uber would end up selling their technology to Aurora, who is — as of 2025 — starting to run autonomous trucking routes.
But yeah, Uber was hellbent on having self-driving technology. They famously partnered with CMU to create an advanced research lab near the campus, and then proceeded to poach 50 people who were involved with CMU’s autonomous vehicles program.
By this point, everyone wanted a piece of the action. Apple poached a Waymo employee. Tesla poached Andrej Karpathy from OpenAI for their own autonomous vehicles efforts. Traditional car companies started feeling like they were falling behind and started acquiring startups, like GM acquiring Cruise for a billion dollars.
At the peak of this jostling, Google accused Uber of coordinating with Anthony Levandowski to steal a bunch of company secrets. Google had proof that Levandowski had downloaded tens of thousands of documents just before leaving Google. A civil suit and a criminal case followed. At the end of it, Waymo got $245 million in Uber stock in the civil trial, and at the end of the criminal trial Anthony Levandowski was sentenced to 18 months in prison.
This is what’s exciting about Meta poaching employees from OpenAI. All of these corporations slugging it out in the news while we get to eat some popcorn and watch. Does past history foreshadow what will happen? Once people get their $100 million payouts, will they feel any pressure to continue working at Meta? Will these companies start accusing each other of steealing industry secrets? No way, this is AI. Something weirder is going to happen. And I’m excited to see what it is.
The self-driving car industry existed before all of the events in this post, but a large portion was driven by researchers trying to meet DARPA contracts and challenges. At a robotics conference in the early 2010s, I got a chance to ride in a self-driving car project run by a university team. It drove close to a large branch that was overhanging the unpaved parking lot, and brushed against it. It was a bit unnerving, especially since nobody was behind the driver’s seat. It felt like being in a car that was rolling uncontrollably down a hill. A researcher explained to me that it didn’t do great with fine obstacles like that branch, but that they had to route it near the branch because otherwise it would have too much trouble making a turn so that it could return to start. My reaction was some version of, “wow, that’s all very impressive. So remind me again, how do you stop this in an emergency?” But since it moved at all and stayed on its course, this was one of the better efforts I had heard of at the time.
This name will come up again later.