New AI lab Core Automation 'nerdsniped' researchers from Anthropic, Google DeepMind

Ah, the AI talent wars — where the strategy is 'throw money at the problem' and the combatants are laballed like rare Pokémon. The latest contender, Core Automation, has 'nerdsniped' a senior researcher from Anthropic's core technical staff and another from Google DeepMind's reasoning team. Yes, 'nerdsniped' is the official Business Insider verb — think of it as industrial-strength poaching dressed up in internet slang. The start-up's founder, a former OpenAI boffin, has apparently discovered that the fastest way to become a contender is to pay someone who was actually building the leading models slightly more than their old lab did. Marvel.

And this isn't just anyone from the old guard — Anil Kumar, a key architect of Claude's 'character' and safety guardrails, has defected. Also on the trophy list is DeepMind's team lead on CoRAG, a technique that's apparently meant to make models check their own reasoning, because we clearly need the reasoning-checked faster without anybody checking the checkers. The lab bills itself as 'maximum focus, minimum friction', which is corporate-speak for 'we're not going to bother with all that expensive safety infrastructure.' Three cheers for velocity over rigour.

The deeper comedy is the sheer disposability of it all. Anthropic and DeepMind spent years cultivating these stars; Core Automation sacked them in one press cycle. The newcomer hasn't produced a single commercial result yet, but who cares when the hiring announcement gets you a free Business Insider headline? The lab world is now running on the same logic as football transfers — hype now, performance… well, maybe later. The only missing piece is jersey sponsorship and a chunky wage cap to keep things honest.