Meta is rattling OpenAI with an aggressive hiring spree. In just one week, it pulled in eight researchers from OpenAI.
The social media giant has stepped up recruitment for its Superintelligence team. Recent hires include Shengjia Zhao, who worked on GPT-4, Jiahui Yu, former head of OpenAI’s Perception team, Shuchao Bi, who managed multimodal models; and Hongyu Ren, who led training for the o3-mini, o1-mini, and o1 models, where he was a foundational contributor. He also worked on an upcoming open-source model.
Trapit Bansal, a foundational contributor to OpenAI’s first reasoning model, o1 and a key figure in reinforcement learning, has also joined Meta’s AI superintelligence team.
Previously, the company hired three researchers, Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai, who were responsible for setting up OpenAI’s Zurich office last year.
The Wired quoted Meta’s chief research officer, Mark Chen, as saying in a Slack memo to their staff that the departures felt “as if someone has broken into our home and stolen something.”
Chen added that the leadership team, including CEO Sam Altman, is working “around the clock” to engage with employees who have received competing offers from Meta.
Amid the hiring tension, OpenAI is also taking a collective pause to address the overwork culture. The company has announced a week off for all employees. “This has been planned for quite some time. I won’t speak for anyone but myself, but I’m super happy for the company to be closed for the week,” said OpenAI’s Adam Goldberg in a post on X.
Goldberg said that taking individual time off can feel like you’re missing out on what’s happening at work, but when the entire company shuts down for a week, it allows everyone to truly unwind without that sense of missing something. “It’s been an amazing first half of the year, now with Q2 close, we recharge and then get back to the grind for a big 2H and beyond.”
Meta is a threat to OpenAI
In a recent podcast with his brother, Jack Altman, Sam Altman said that Meta views OpenAI as its primary competitor. He also mentioned that Meta is offering extremely high compensation packages to OpenAI employees, including signing bonuses of $100 million or more.
Meta CEO Mark Zuckerberg is personally recruiting for a new “superintelligence” team, offering hefty payouts to top AI researchers across the industry.
Just last week, the tech giant announced a $14.3 billion investment in Scale AI and brought on Scale’s CEO, Alexandr Wang, to join the Superintelligence team.
Meta, too, like OpenAI, uses the term “superintelligence” rather than AGI, even though the tech and AI giants have yet to achieve it.
That said, in a recent blog post, Altman wrote that the arrival of superintelligence could be just a few thousand days away. “It is possible that we will have superintelligence in a few thousand days (!). It may take longer, but I’m confident we’ll get there.”
Altman added that many people in Meta’s AI team are trying to copy OpenAI. He also said that while OpenAI began as a deep research lab, it has recently had to evolve into a commercial tech product organisation, especially after launching ChatGPT.
Altman stated that while many around the world view ChatGPT as a substitute for Google, at Meta, it is seen more as a replacement for Facebook. He also mentioned that users are increasingly turning to ChatGPT instead of traditional social media platforms.
Altman suggests it’s not just about how much time people spend online, but how they feel afterwards. “People like doomscrolling on the internet. It may feel good at the moment, but it’s making you feel worse,” he said.
By contrast, he says users often walk away from ChatGPT with a more positive outlook.“A thing that we’re very proud of is when people talk about ChatGPT, they’re like, ‘I actually like myself better.’”
Based on these observations, OpenAI has made its latest acquisition. The AI startup has acquired Crossing Minds, a company specialising in infrastructure for personalised recommendations and real-time machine learning.
This move is likely to help OpenAI capture more user attention and time, as it works to make ChatGPT more personalised based on individual interests. Notably, ChatGPT now also features a memory function that allows it to recall past conversations.
OpenAI has recently made ChatGPT’s memory feature available to free users, a function that was earlier limited to Plus and Pro subscribers. As of June 3, 2025, this “lightweight” version enables ChatGPT to recall recent conversations, allowing it to deliver more personalised and context-aware responses while reducing the need for users to repeat themselves.
Founded in early 2017 by Sebastian Thrun, Emile Contal, and Alexandre Robicquet, Crossing Minds was started to build AI that understands people on a deeper level, not just what they might do next, but what they care about over time.
The company has developed tools that integrate with real-time systems and complex infrastructure, and more recently, it has explored new methods for handling information using large language models.
“Truly understanding people’s intents is the foundation of outstanding AI assistants,” the Crossing Minds team said in a statement. “Not merely predicting next actions, but learning long-term preferences, and what makes each of us unique.”
A few months ago, reports surfaced that OpenAI is working on a social media platform similar to X. The platform is expected to combine ChatGPT’s advanced AI features, particularly its image generation tools, with a personalised content feed.
Soon after, Meta released the first version of its standalone Meta AI app. It is built to provide a personalised assistant experience across devices, including Meta’s AI-powered Ray-Ban glasses and the web. The app features voice-based interaction, contextual memory, and a Discover feed that showcases how users are engaging with AI.
This seems to be just the beginning of the war between Meta and OpenAI. The latter is also planning to release an open-source model next month.
“Sorry to hype, but having a few friends at OpenAI makes it hard not to hear how wild their open-source model dropping next month is. It won’t run on a phone, but let’s just say it edges out one of the models in the ChatGPT dropdown. Meta’s gonna need some time to catch up,” said Yuchen Jin, cofounder of Hyperbolic.
OpenAI’s message to Meta seems clear: catch us if you can.
The post Inside the Great AI Talent Heist of 2025 appeared first on Analytics India Magazine.
