Core Automation nerdsniped researchers from Anthropic and Google DeepMind, and the AI market is paying attention
The artificial intelligence market is boiling over right now, and the latest move getting everyone talking involves heavyweight talent leaving giants like Anthropic and Google DeepMind to jump into something completely new.
Core Automation is the AI startup of the moment, and it showed up making noise.
Founded by Jerry Tworek, former vice president at OpenAI, the company introduced itself to the world with a bold pitch: build the most automated AI lab on the planet. Tworek lists himself as CEO and co-founder of Core Automation in his X bio, making it clear he is not just passing through.
But what really grabbed attention was not just the startup’s mission — it was who agreed to drop everything to be part of it.
Researchers who were comfortably settled at some of the best AI labs in the world simply turned the page, and the reason even has a name: nerdsniped 🧠
For anyone unfamiliar with the term, nerdsniped is that feeling of being completely hooked by a problem or idea so interesting that you just cannot resist, even if it means changing course entirely.
And that is exactly what happened here.
Where these people came from and why it matters
When an AI startup manages to attract researchers who were working at places like Anthropic and Google DeepMind, that is not a small detail. These two labs are among the most respected and well-funded in the industry. Anthropic, founded by former OpenAI members, is known for its deep work on AI safety and for the Claude models, which rank among the most advanced available today. Google DeepMind, on the other hand, is responsible for breakthroughs that shaped the entire field, from AlphaFold to Gemini. Leaving one of these places is not a simple decision — it is a clear bet on something that, in these people’s view, is more worth pursuing.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Rohan Anil, a researcher at Anthropic who had also previously worked at Google DeepMind, publicly confirmed on X that he left the company after Tworek nerdsniped him. In his own words, Anthropic was one of the best places to work as a researcher, but Core Automation’s pitch was compelling enough to make him change his mind.
Anmol Gulati, a research scientist at Google DeepMind who worked directly on Gemini, also announced he was heading off to something new alongside exceptional people. Gulati went further and shared his view on the current state of the field, saying he increasingly felt that the current research paradigm — based on scaling models, data, and static deployments — would not be enough to take the field where it needs to go. In his view, the next phase comes from something different: new learning algorithms, architectures that go beyond the current stack, and systems that automate the very process of building.
This kind of statement carries weight because it comes from someone who was inside one of Google’s most ambitious projects, Gemini. When a researcher of that caliber says the current path has limitations, it is worth paying attention to what he is proposing as an alternative.
The team forming at Core Automation
The roster of names boarding the Core Automation ship goes well beyond Anil and Gulati. Joanne Jang, who served as general manager at OpenAI from December 2021 to April 2026, updated her X bio with the phrase trying to automate myself @coreautoai. That kind of positioning shows the startup’s culture is already locked in from day one: automation is not just a product — it is a philosophy that runs through everything.
Other names who confirmed their involvement include Ehsan Amid and Avery Lamp, both former Google DeepMind employees, along with Julia Villagra, who was head of people at OpenAI, and Sai Surya Duvvuri, a former research intern at Google and Meta. Villagra’s presence is particularly interesting because it signals that Core Automation is not just thinking about pure research but also about how to build a strong organization from the ground up, with culture and people management being a priority from the start.
On the company’s website, Core Automation describes its team as people who helped build frontier models and influential architectures. That is not an empty claim, considering the track record of each person joining the project. We are talking about professionals who contributed directly to systems like Gemini, Claude, and OpenAI’s models — in other words, the cream of the crop in global AI research.
What becomes clear is that Core Automation is not just hiring impressive resumes. It is bringing together people who were already at the top of the AI research chain and who, despite that, felt that itch to go do something different. This kind of movement usually happens when an idea is genuinely disruptive — when it proposes a path that the major players are not yet willing to take, whether because of bureaucracy, because their focus is on other products, or simply because of the sheer size of the structures they have to carry.
What is Core Automation and what does it want to build
Core Automation positions itself as an AI startup with a very specific and ambitious goal: to create the most automated artificial intelligence lab in the world. In its first post on X, the company made the central objective clear: systems that optimize and automate work, starting with research itself.
That means developing systems where AI research itself is conducted, in large part, by other AIs. It is a concept that blends systems engineering, research in large language models, process automation, and a very particular vision of how the future of science can be accelerated. Jerry Tworek, who served as vice president at OpenAI before creating Core, clearly carries with him a deep understanding of how these systems work and, more importantly, where the bottlenecks are that still hold back progress in the field.
The idea of automating AI research is not new, but the way Core Automation is approaching it seems to have something different going on. Instead of using automation merely as a support tool for human researchers, the proposal goes further — building infrastructure where experimentation, evaluation, and iteration cycles happen at a speed that humans simply could not sustain on their own. This has enormous implications for the pace of new model development, for identifying flaws, and for discovering approaches that have not even been tried yet.
It is exciting territory and, at the same time, full of open questions about how it would actually work in practice. But the combination of a clear vision and an extremely high-caliber technical team is exactly the kind of foundation that gives credibility to a proposal this ambitious.
The nerdsniped effect and what it reveals about the industry
The term nerdsniped went viral in the tech community because it describes with precision something that anyone passionate about technology has felt before: that moment when a problem or idea hooks you so completely that you simply cannot think about anything else. It is almost an intellectual compulsion. And when it happens on a collective level — when multiple highly qualified people have the same reaction to the same project — it is a sign that something genuinely different is being proposed.
Core Automation appears to have triggered exactly that response in people who, by all accounts, had every reason to stay where they were.
This phenomenon also reveals something about the current state of the artificial intelligence industry. Even with well-funded labs, sky-high salaries, and globally impactful projects, there is still a significant group of researchers who feel that the most interesting problems are not being tackled the right way. There is a restlessness in the air — a sense that the field has not yet arrived where it needs to be and that existing structures, no matter how powerful, may not be the best environment for solving certain questions. That restlessness is fuel for startups like Core Automation, which show up at exactly the right moment to channel that energy.
On top of that, the nerdsniped effect has a very concrete practical consequence: it attracts top-tier talent without the company needing to compete solely on financial terms. In an industry where salaries are already through the roof and big-lab perks are tough to beat, the ability to present a genuinely fascinating problem is a real competitive advantage. Core Automation figured that out, and the result is visible in the composition of the team being assembled. 🚀
The trend of talent leaving Big Tech for AI startups
The migration of renowned researchers from major labs to startups is not something Core Automation invented. This is a movement that has been intensifying over the past few years and reflects an important shift in industry dynamics.
A recent and notable example is Yann LeCun, who was chief AI scientist at Meta and left to found his own startup, Advanced Machine Intelligence Labs, known as AMI Labs. AMI Labs focuses on developing world models — AI systems capable of understanding and reflecting the real world more faithfully. AMI Labs’ approach diverges from Meta’s focus on developing models geared toward commercial purposes and scale, which reinforces the idea that top researchers frequently pursue directions that large corporations do not prioritize.
Over the past year, big tech companies waged all-out wars for AI talent, offering multibillion-dollar acquihires and massive compensation packages. But startups also proved to be active players in this battle, offering competitive salaries, generous equity packages, and something that large companies have a harder time delivering: direct impact and a real sense of ownership over what is being built.
Shawn Thorne, managing director at executive recruiting firm True Search, noted that base salaries at startups climbed rapidly as they competed to attract AI talent. According to him, equity is the major factor that helps offset the opportunity cost for top-tier researchers and engineers who might otherwise choose to start their own ventures.
To make the proposition even more appealing, startups also offer additional incentives like co-founder titles, access to compute resources at scale, and dedicated time for independent research. These perks might seem subtle at first glance, but for a researcher who wants intellectual freedom while also being part of something with massive impact potential, they make all the difference.
What this signals for the future of AI research
The creation of Core Automation and the rapid concentration of talent around it raise important questions about where the field is headed. If researchers of the highest caliber are saying that the current paradigm of scaling models and data has limitations, and they are putting their careers behind a different approach, that is an indicator that deserves attention.
The proposal to automate research itself might sound abstract, but think about the practical impact: if an AI can run experiments, analyze results, and propose new avenues of investigation autonomously, the pace of innovation could accelerate exponentially. Instead of relying exclusively on the time and cognitive capacity of human researchers, discovery cycles could become continuous, operating at scales and speeds that are simply not feasible today.
Of course, this brings enormous challenges. Questions about quality, safety, interpretability, and control of these automated systems need to be answered. But the fact that people with direct experience building frontier models like Claude and Gemini are dedicating themselves to this problem suggests there is a solid technical foundation behind the ambition.
The move Core Automation is making is one of those worth watching closely. Not just because of what the company promises to deliver, but because of what the composition of its team already says about the direction the artificial intelligence field may be taking. When the best talent in the industry starts pointing toward the same place, it is usually worth paying attention. 👀
