A tiny AI agent from a London startup just beat Anthropic and OpenAI at reading scientific papers

Key points

• Inherent, a London AI startup founded by four Google DeepMind alumni, raised a $50 million seed round before releasing its first public results.
• Its AI agent, Faraday, beat Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5 at independently replicating published scientific research without being given the answers first.
• Faraday runs on Qwen 3.6, a model with 27 billion parameters, a proxy for size and training cost, making it far smaller than the frontier models it outscored.
• Inherent uses reinforcement learning, a training method that rewards good outcomes rather than spelling out fixed rules, to teach Faraday what researchers call "research taste."
• The company plans to grow from 12 to roughly 20 to 25 employees by the end of 2025.

A small London lab is making a loud claim. Inherent, whose four founders all left Google DeepMind to start it, says its AI agent has beaten some of the most capable AI systems on the market at a standard scientific task, using a fraction of the computing muscle.

The company only came out of stealth a few weeks ago, announcing a $50 million seed round. Now it is sharing the first real look at what it has built.

What did Faraday actually do?

Faraday was asked to replicate the findings of published scientific papers on its own, without being told the correct answers in advance. That means reading the paper, deciding what experiments to run, and arriving at the same conclusions the original researchers did.

It beat both Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5, two of the largest and best-funded AI models available, at this task. First reported by TechCrunch, the result draws attention because of how Faraday pulled it off.

Faraday runs on Qwen 3.6, a model with 27 billion parameters. Parameters are the internal numeric settings that define how a model thinks; more parameters generally mean a bigger, more expensive model to train and run. Claude Opus 4.8 and GPT-5.5 are widely understood to operate at a far larger scale.

Why does paper replication matter?

It is a standard starting exercise for new scientists. "Many PhD students actually start by doing this," said Edward Hughes, Inherent's cofounder and chief scientist. The point is not just to get the right answer but to show you understand the process well enough to rebuild it from scratch.

Inherent's bar was higher still. Beyond accuracy, it wanted Faraday to show what Hughes calls "research taste", an instinct for which experiments are worth running and how to design them well. Teaching something that intangible required a specific approach.

The team used reinforcement learning, a training method that rewards an AI for producing good outcomes rather than giving it a rule book to follow. The bet is that reward-based learning produces an agent that generalises across many scientific fields, not just the ones it trained on.

Faraday also uses OpenAI's GPT-5.5 Codex for coding tasks, the same way a working scientist reaches for existing software rather than writing everything from scratch.

What does this mean for the future of AI research tools?

Inherent's longer goal is an AI that does not just verify old science but discovers new things entirely. Replication is the foundation, not the destination.

Hughes describes his ideal agent as a teammate who "got curious about this, and went off and did these experiments" and then asks what you think of the results. That collaborative quality, he says, is what Inherent is trying to build in.

The company's 12 staff all work in person from King's Cross in London. Hughes has publicly called for the UK to scrap "garden leave", a common British employment clause that can stop departing workers from joining or founding a rival for months after they quit. He says the rule slows UK AI startups relative to their US rivals. Inherent plans to hire up to 25 people before the year is out.