Frontier talent poached
- A startup called Core Automation has recruited researchers from Anthropic and Google DeepMind. - The hires signal elite lab researchers are mobile and willing to join well‑funded startups. - That portability tightens competition for scarce research talent and expands viable career targets beyond incumbents. (businessinsider.com)
Core Automation, a new artificial intelligence startup, has pulled researchers from Anthropic and Google DeepMind into a lab led by former OpenAI executive Jerry Tworek. (businessinsider.com) Core Automation introduced itself on X on Tuesday, April 21, saying it is “building the world’s most automated AI lab” and wants to automate work “starting with research itself.” Tworek lists himself as Core Automation’s chief executive and cofounder. (finance.yahoo.com) One of the first public defections came from Rohan Anil, who had worked at both Google DeepMind and Anthropic. He said on X that he left Anthropic after Tworek “nerdsniped” him into joining Core Automation. (africa.businessinsider.com) The move lands in a market where a small number of researchers can switch the trajectory of a lab. Radical Ventures wrote on April 20 that “neolabs” backed like research institutions are widening the set of employers competing for frontier artificial intelligence talent. (radical.vc) That competition has been getting more expensive. USA Today reported in May 2025 that Google DeepMind had offered some top researchers about $20 million a year, showing how far major labs will go to keep or win scarce talent. (usatoday.com) Core Automation is also pitching a different research bet than simply training ever-bigger models on ever-more data. Coverage of the launch said the company is pursuing learning methods beyond standard pretraining and reinforcement learning, while trying to use software tools and agents to automate more of the research process. (the-decoder.com) That pitch helps explain why a startup can recruit from incumbents with far larger balance sheets. Early-stage labs can offer equity, senior titles, direct access to computing resources, and room to shape a research agenda instead of fitting into an established one. (europesays.com) Anthropic, Google DeepMind, and OpenAI still have the cash, chips, and existing models that startups usually lack. Core Automation’s opening week shows that those advantages no longer guarantee that top researchers will stay put. (businessinsider.com)
