{"id":651232,"date":"2026-08-22T20:29:08","date_gmt":"2026-08-22T20:29:08","guid":{"rendered":"https:\/\/www.europesays.com\/ie\/651232\/"},"modified":"2026-08-22T20:29:08","modified_gmt":"2026-08-22T20:29:08","slug":"inherent-founded-by-deepmind-alumni-says-its-ai-teammate-just-outperformed-anthropic-and-openai-at-replicating-research","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ie\/651232\/","title":{"rendered":"Inherent, founded by DeepMind alumni, says its AI ‘teammate’ just outperformed Anthropic and OpenAI at replicating research"},"content":{"rendered":"

Inherent<\/a>, a London AI lab founded by Google DeepMind alumni<\/a>, says its AI agent just outperformed much larger models from Anthropic and OpenAI using a fraction of the size.<\/p>\n

Of all the startups launched by Google DeepMind alumni, Inherent has gotten relatively little attention. But while better-funded rivals have yet to show the world anything concrete, the London-based team is starting to share what it\u2019s been building.<\/p>\n

Just weeks after emerging from stealth with a $50 million seed round<\/a>, the British startup says its newly released AI agent, Faraday<\/a>, has outperformed<\/a> larger, better-known models at a specific task: independently reproducing the findings of published scientific papers without being told the answer in advance.<\/p>\n

That may sound like a mere party trick given Inherent\u2019s much loftier goal \u2014 building AI that can discover new scientific knowledge and not just verify old results. But paper replication is a standard training exercise for human scientists, too, cofounder and chief scientist Edward Hughes said. \u201cMany PhD students actually start by doing this.\u201d<\/p>\n

Beating other AI systems at the task wasn\u2019t the point, Hughes told TechCrunch; how they got there was. \u201cWhat was most interesting to us about this was not so much the result of beating those frontier agents \u2014 which of course we liked \u2014 but was actually the way we went about building this.\u201d<\/p>\n

Here\u2019s the part that should catch an investor\u2019s eye: measured against Anthropic\u2019s Claude Opus 4.8 and OpenAI\u2019s GPT-5.5 \u2014 both much larger, frontier-scale systems \u2014 Faraday runs on a comparatively tiny model called Qwen 3.6 that has just 27 billion parameters. (Roughly speaking, \u201cparameters\u201d is a proxy for a model\u2019s size and, typically, its training costs, as well.) Inherent\u2019s bar for success was also higher than simply accuracy. Beyond replicating results, it wanted Faraday to demonstrate \u201cresearch taste\u201d \u2014 an instinct for what experiments are worth running and how to design them well.<\/p>\n

Teaching something as intangible as taste is hard, which is where reinforcement learning<\/a> comes in. It\u2019s a training method that rewards an AI system for good outcomes rather than spelling out rules for it to follow. Rather than training its agents primarily on the study of how science itself is conducted, Inherent leans on this reward-based approach, betting it will generalize better to its longer-term goal of agents capable of contributing across many scientific fields.<\/p>\n

\u201cWe\u2019re always guided by that north star of building an AI scientist agent and imbuing our agents with taste,\u201d Hughes said. That focus has also shaped what Inherent chooses not to build. Rather than developing its own coding tool, it had Faraday use OpenAI\u2019s GPT-5.5 Codex instead, much the way human scientists lean on existing software rather than building everything themselves, according to the company.<\/p>\n

Inherent is also trying to avoid building agents that simply tell users what they want to hear. Instead, Hughes said, the goal is modeled on his favorite kind of teammate \u2014 the kind who comes back and says: \u201cI got curious about this, and I went off and I did these experiments. What do you think of these results?\u201d<\/p>\n

That collaborative instinct extends to how Inherent operates as a company. Its dozen employees all work in person out of an office in King\u2019s Cross \u2014 the once-rundown London neighborhood that Google DeepMind\u2019s presence helped turn into one of the world\u2019s top AI hubs<\/a>. \u201cWe believe that London is the place to be,\u201d Hughes said.<\/p>\n

Hughes is bullish on London\u2019s density of AI talent, but he has also added his voice<\/a> to calls to end \u201cgarden leave\u201d \u2014 the practice, common in the U.K., of barring departing employees from joining or starting a rival company for months after they resign. It\u2019s a restriction American researchers generally don\u2019t face<\/a>, giving U.S. startups a head start on hiring talent who\u2019ve left a prior role. \u201cThis is a personal view rather than a company view, but I was affected by the garden leave problem,\u201d he told TechCrunch.<\/p>\n

Hughes eventually got around that constraint and started Inherent alongside two other DeepMind alumni and a fourth cofounder. The startup isn\u2019t slowing down either. It plans to grow its headcount to \u201cabout 20 to 25\u201d by the end of the year. Given its ambitions in world models<\/a> as well, and with Demis Hassabis\u2019s new role leaving some DeepMind staff unsettled<\/a>, Inherent\u2019s hiring push could make it an appealing landing spot for DeepMind employees weighing a move.<\/p>\n

Pictured from left to right: Inherent co-founders Louis Kirsch, Kaloyan Aleksiev, Tantum Collins and Edward Hughes.<\/p>\n

When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/p>\n","protected":false},"excerpt":{"rendered":"Inherent, a London AI lab founded by Google DeepMind alumni, says its AI agent just outperformed much larger…\n","protected":false},"author":2,"featured_media":651233,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","_share_on_mastodon":"0"},"categories":[261],"tags":[291,289,290,18,20321,19,17,130,82],"class_list":["post-651232","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-eire","tag-google-deepmind","tag-ie","tag-ireland","tag-london","tag-technology"],"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@ie\/117141056064813271","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/651232","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/comments?post=651232"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/651232\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media\/651233"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media?parent=651232"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/categories?post=651232"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/tags?post=651232"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}