EN
Back to the archive

The archive · AI & Models · Product decision · 2025–2026

Inherent bets AI-native science needs new labs: ex-DeepMind team raises $50M

Ex-DeepMind researchers' London lab Inherent builds Faraday, an AI system for open-ended discovery; $50M seed led by Index Ventures and Radical Ventures.

Inherent

The betMachines capable of open-ended exploration can catalyze new inventions — science must be rebuilt around humans and self-improving AI, not AI grafted onto old methods.Building

What the business is

Inherent is a London AI lab building Faraday, an AI system that pairs human researchers with self-improving AI to pursue open-ended scientific discovery; it calls the goal 'AI-native science' and aims to write its playbook.

Starting capital$50 million seed, led by Index Ventures with Radical Ventures (announced May 2026)

How it started

Tantum Collins and Edward Hughes first met while researching cooperative AI at DeepMind; Collins later handled AI policy at the Biden White House. With fellow DeepMind researcher Louis Kirsch and infrastructure specialist Kaloyan Aleksiev (Reka AI, Microsoft), they founded Inherent in stealth in 2025. Faraday, named after the 19th-century scientist Michael Faraday, is the system they set out to build.

What happened

In May 2026 Inherent emerged from stealth with a $50M seed led by Index Ventures, with Radical Ventures participating. Matt Clifford CBE — Entrepreneurs First co-founder and former UK government AI adviser — joined as an advisor, and early hires and angels include technical-safety and democratic-AI specialists. The company is incorporated as a public benefit corporation; Index Ventures calls it a 'recursive company' where organizational design and research feed each other.

No ending yet — it is still running.

Background

Inherent is a London-based AI lab founded by former DeepMind researchers Tantum Collins, Edward Hughes and Louis Kirsch, together with Kaloyan Aleksiev, who came from Reka AI and Microsoft. It emerged from stealth in May 2026 with a $50M seed led by Index Ventures with Radical Ventures participating, and describes its mission as writing 'the playbook for AI-native science': a system called Faraday, named after Michael Faraday, in which human researchers and self-improving AI work together on the hardest problems in science.

The founding bet is that most AI gives good answers but cannot figure out which questions are worth asking, and that machines capable of open-ended exploration can catalyze a new wave of inventions. Inherent's manifesto draws an analogy with steam-driven factories: electrification only paid off once the factory was redesigned, and AI-native science similarly needs institutions built from scratch. The company applies recursive self-improvement at the level of the whole lab and is a public benefit corporation, with a societal-benefit board to be announced.

Collins and Hughes first met researching cooperative AI at DeepMind, and Collins worked on AI policy at the Biden White House before co-founding. Matt Clifford CBE, co-founder of Entrepreneurs First and a former UK government AI adviser, is an advisor, and early hires and angels include technical-safety and democratic-AI specialists. The lab says how its humans and AI agents collaborate day to day is itself part of the research that informs Faraday.

What has to be true

  • DeepMind alumni saw frontier AI's limit first-hand: systems excel at convergent, well-specified problems but flounder at open-ended discovery — deciding which questions are worth asking at all.
  • Treating organizational design as a research problem is a differentiation frontier labs cannot buy: the way Inherent runs becomes the moat that shapes the system it is building.
  • A $50M seed led by Index Ventures with Radical Ventures, plus advisers like Matt Clifford, is institutional conviction that AI-native science can become a category of its own.
  • Incorporating as a public benefit corporation with technical-safety specialists among early hires answers the governance objection that frontier AI labs usually face later.

What can be applied

Bold AI bets can be organizational before technical: Inherent made human-agent collaboration part of the product itself — investors funded the thesis, but real discoveries are the test.

Aftermath

As of September 2, 2026, Inherent is out of stealth and still building: no model, product or scientific result is public yet, and its manifesto describes Faraday as in development with recruitment open. The $50M seed announced in May 2026 funds early work on AI scientists for open-ended research. The company competes in a crowded space — DeepMind itself has AlphaFold and other scientific tools, and many startups target the same AI-plus-research intersection — so the open question is whether Faraday can produce genuine breakthroughs, not incremental productivity gains.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases