Analysis
Ineffable Intelligence, the superintelligence lab founded by former Google DeepMind reinforcement-learning chief David Silver, has named six of its senior staff as cofounders, Fortune reported Monday morning. None of the six were at the company when it was incorporated in November 2025. All of them now carry the title.
The six, and what each one brings:
- Chris Apps -- former technical program and research delivery lead at Google DeepMind, previously Barclays and Deloitte. Runs what Ineffable calls mission acceleration.
- Wojciech Czarnecki -- ex-DeepMind researcher on multi-agent systems and AlphaStar. Oversees the science team.
- Lasse Espeholt -- built Google's MetNet weather-prediction models. Leads compute, infrastructure and engineering strategy.
- Junhyuk Oh -- DeepMind's former reinforcement-learning lead on AlphaStar, and a pioneer of AI-discovered RL algorithms. Runs reinforcement learning at Ineffable.
- Alexandre Laterre -- former head of research at InstaDeep, the London lab BioNTech acquired in 2023. Oversees research engineering.
- Heather Gorham -- former venture partner at Flying Fish, who approached Silver in early 2024. Handles compute procurement, fundraising and operations.
“- Alexandre Laterre -- former head of research at InstaDeep, the London lab BioNTech acquired in 2023.”
How a lab gets to $5.1 billion with no product
Silver is the researcher behind AlphaGo and AlphaZero, and the thesis at Ineffable is a direct extension of that work: build a "superlearner" trained primarily through reinforcement learning rather than next-token prediction over human text. He became a company director in January 2026 after leaving DeepMind. George Rose, the Canadian entrepreneur who actually incorporated the entity, is now listed as a founding advisor rather than a cofounder -- a detail worth sitting with, given that today's announcement adds six cofounders while reclassifying the person who filed the paperwork.
In April 2026 the company came out of stealth with $1.1 billion at a $5.1 billion valuation, co-led by Sequoia and Lightspeed with Nvidia, DST Global, Index, Google and the UK's Sovereign AI Fund participating, per CNBC. It was the largest seed round in European history. TechCrunch's account of the round noted the company had no product and no revenue at the time. Five months later, that has not changed publicly.
The comparison set
The closest comparables are American:
- Thinking Machines -- roughly $2B seed at $12B, now [in talks with Accel](/pulse/thinking-machines-1-billion-round-40-billion-valuation-2026) for $1B at $40B: founded 2025 by former OpenAI CTO Mira Murati. Ships Tinker, a model-adaptation platform, at a run rate above $100 million.
- Safe Superintelligence: Ilya Sutskever's lab, raised on the same founder-and-thesis basis with no product disclosed.
- Ineffable Intelligence -- $1.1B at $5.1B: the only one of the three based outside the US, with a sovereign fund on the cap table.
What separates Ineffable is not just geography. Its research approach deliberately breaks from the transformer-scaling orthodoxy that every other lab is spending against, which means its compute bill buys a different experiment rather than a bigger version of the same one.
The counterweight
Cofounder titles awarded ten months after incorporation are, at minimum, an unusual instrument. They are typically used to close a hiring gap that cash and standard option grants could not -- which tells you what the market for DeepMind's reinforcement-learning bench costs right now. They also dilute the meaning of the title internally, and they raise a fair question about what founder-class equity and vesting look like for six people who joined a company already valued in the billions. Fortune's account does not disclose those terms, and Ineffable has not published them.
The harder risk is scientific. Reinforcement learning without large-scale human-text pretraining is a genuinely different bet, and it has not yet produced a general-purpose system competitive with frontier LLMs. If it works, the $5.1 billion mark will look cheap. If it takes five years, the lab burns Nvidia-priced compute the entire time with no revenue to offset it, and the next round gets priced by whoever is still willing to fund pure research.
Silver's team now has to convert names into a demonstrable result. The first real signal will be whether Ineffable publishes anything -- a paper, a benchmark, a system card -- before it needs to raise again.
