LONDON — Inherent, a London-based AI lab founded by Google DeepMind alumni, said its newly released agent, Faraday, outperformed Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5 at independently reproducing the findings of published scientific papers — while running on a model with just 27 billion parameters, a fraction of its rivals’ scale.
The claim, made public weeks after Inherent emerged from stealth with a $50 million seed round, positions the small startup as a serious contender in the race to build AI that can conduct scientific research autonomously. But cofounder and chief scientist Edward Hughes emphasized that the benchmark result was secondary to the method behind it.
Also read: Harvard’s $699 Startup Bootcamp Deploys AI Avatars of Its Instructors
“What was most interesting to us about this was not so much the result of beating those frontier agents — which of course we liked — but was actually the way we went about building this,” Hughes told TechCrunch.
Replication as a training ground for AI scientists
Inherent’s stated goal is far more ambitious than verifying old results: it aims to build an “AI scientist” capable of discovering new knowledge. However, the company sees paper replication as a foundational exercise, much like a PhD student’s early training.
Also read: Frontier AI Labs Fall Short on Public Rogue-Model Containment Plans, Study Finds
“Many PhD students actually start by doing this,” Hughes said. The task requires an agent to read a paper, design experiments, run them, and arrive at the same conclusions — without being told the answer in advance.
According to Inherent, Faraday achieved this using a reinforcement learning approach that rewards good outcomes rather than following explicit rules. The company deliberately avoided training its agents primarily on the study of science itself, betting that reward-based learning will generalize better to its long-term goal of contributing across multiple scientific fields.
“We’re always guided by that north star of building an AI scientist agent and imbuing our agents with taste,” Hughes said, referring to the instinct for choosing worthwhile experiments and designing them well.
A different approach to building AI
Inherent’s strategy extends to what it chooses not to build. Rather than developing its own coding tool, Faraday used OpenAI’s GPT-5.5 Codex — much as human scientists rely on existing software rather than reinventing it.
The company is also trying to avoid creating agents that simply tell users what they want to hear. Hughes said the goal is modeled on his favorite kind of teammate: one who returns and says, “I got curious about this, and I went off and I did these experiments. What do you think of these results?”
That collaborative ethos is reflected in Inherent’s operations. Its dozen employees work in person from an office in King’s Cross, the London neighborhood that Google DeepMind’s presence helped transform into a major AI hub. “We believe that London is the place to be,” Hughes said.
What this means for the AI research sector
The announcement adds a new data point to the ongoing debate about whether scale alone determines AI capability. While frontier labs continue to build ever-larger models, Inherent’s results suggest that focused training on specialized tasks — and a relatively small model — can achieve competitive performance in narrow domains.
Industry observers will likely watch whether Inherent can extend this approach to broader scientific discovery, a far harder problem. The company also plans to grow its headcount to roughly 20–25 employees by the end of the year, and with Demis Hassabis’s new role leaving some DeepMind staff unsettled, Inherent’s hiring push could attract talent from its former parent.
Hughes has also publicly called for an end to “garden leave” in the U.K., a practice that bars departing employees from joining or starting a rival company for months after resignation. He said he was personally affected by the restriction, which he believes gives U.S. startups an unfair advantage in hiring. “This is a personal view rather than a company view, but I was affected by the garden leave problem,” he told TechCrunch.
For now, Inherent’s focus remains on its north star: building an AI scientist that can contribute meaningfully to research. Whether Faraday’s early success translates into genuine discovery remains an open question — but the startup has made clear it intends to find out.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. The cryptocurrency and AI markets are volatile and uncertain; readers should conduct their own research before making any investment decisions.
Be the first to comment