[{"@context":"https:\/\/schema.org\/","@type":"BlogPosting","@id":"https:\/\/blog.terabox.com\/insights\/benchling-ai-agents-life-sciences-research#BlogPosting","mainEntityOfPage":"https:\/\/blog.terabox.com\/insights\/benchling-ai-agents-life-sciences-research","headline":"Benchling AI: Building Specialized Agents for Life Sciences","name":"Benchling AI: Building Specialized Agents for Life Sciences","description":"\ud83d\udcfa Today’s recommended deep-dive video: https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE Accelerating Biology: How Benchling is Building AI Agents for Life Science R&DThe Foundation of Data-First AgentsArchitecture and the “Scientific” HarnessSkills, SOPs, and the Future of Lab WorkKey TakeawaysQ&A Accelerating Biology: How Benchling is Building... ","datePublished":"2026-07-22","dateModified":"2026-07-22","author":{"@type":"Person","@id":"https:\/\/blog.terabox.com\/author\/flextech-admin\/#Person","name":"flextech-admin","url":"https:\/\/blog.terabox.com\/author\/flextech-admin\/","image":{"@type":"ImageObject","@id":"https:\/\/secure.gravatar.com\/avatar\/ad516503a11cd5ca435acc9bb6523536?s=150&d=mm&r=gforcedefault=1","url":"https:\/\/secure.gravatar.com\/avatar\/ad516503a11cd5ca435acc9bb6523536?s=150&d=mm&r=gforcedefault=1","height":96,"width":96}},"publisher":{"@type":"Organization","name":"terabox","logo":{"@type":"ImageObject","@id":"http:\/\/blog.terabox.com\/wp-content\/uploads\/2021\/11\/logo\u4ea7\u54c1\u540d-\u7ad6\u7248.png","url":"http:\/\/blog.terabox.com\/wp-content\/uploads\/2021\/11\/logo\u4ea7\u54c1\u540d-\u7ad6\u7248.png","width":900,"height":900}},"image":{"@type":"ImageObject","@id":"https:\/\/img.youtube.com\/vi\/RjpTrffSMjE\/maxresdefault.jpg","url":"https:\/\/img.youtube.com\/vi\/RjpTrffSMjE\/maxresdefault.jpg","height":"","width":""},"url":"https:\/\/blog.terabox.com\/insights\/benchling-ai-agents-life-sciences-research","video":{"@context":"http:\/\/schema.org\/","@type":"VideoObject","@id":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE#VideoObject","contentUrl":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE","name":"How to build agents when the smartest AI isn't smart enough","description":"Nick Larus-Stone is the Head of AI at Benchling, the R&D data platform that life science companies use to store and manage their experiments, samples, instruments, and analysis. Benchling has been around for since 2012. In October 2025, it launched Benchling AI, an intelligence layer with a chat interface, backed by an agent, that helps scientists find data, design experiments, and write reports. Nick came to Benchling through its acquisition of Sphinx Bio, the analysis startup he founded. In this conversation, Nick walks through what it takes to build agents for scientific work, and where the playbook from coding agents holds up and where it breaks down.\n\nWe also discuss:\n\u2022 Why Benchling invests so heavily in getting clean data upfront\n\u2022 How they cross-check answers between models to get more out of each one\n\u2022\u00a0Why and how Benchling leans on production traces\n\u2022\u00a0Where AI actually helps science today, and where it still gets stuck\n\u2022\u00a0Why understanding LLMs is closer to biology than software engineering\n\nTimestamps:\n00:00 Intro\n01:22 What Benchling AI is, and the 14-year data platform underneath it\n04:36 Why a decade of structured data is a core advantage\n05:57 The architecture under the hood \n08:28 Similarities and differences compared to a coding harness \n11:14 Benchling\u2019s multi-agent architectures \n14:36 Dealing with verifiable vs non-verifiable tasks \n16:19 Doing evals when clean benchmarks aren\u2019t possible \n18:13 Context engineering: SQL vs. file-based harnesses\n22:11 Memory: agents that create and update their own skills\n25:30 What user education for scientists looks like \n30:33 Why understanding LLMs is closer to biology than software\n33:28 When will agents discover a novel cure for disease? \n44:58 The future of harnesses in science \n48:13 Why fine-tuning on biology hasn't beaten frontier models\n\nReferences:\n\u2022 Agent Skills (Claude Docs): https:\/\/docs.claude.com\/en\/docs\/agents-and-tools\/agent-skills\/overview\n\u2022 Benchling\u2019s Deep Research Agent: https:\/\/www.benchling.com\/blog\/complex-questions-fast-answers-benchling-deep-research\n\u2022\u00a0Claude (Anthropic): https:\/\/www.anthropic.com\/claude\n\u2022 Design of experiments (DOE): https:\/\/en.wikipedia.org\/wiki\/Design_of_experiments\n\u2022\u00a0FDA Investigational New Drug (IND) application: https:\/\/www.fda.gov\/drugs\/types-applications\/investigational-new-drug-ind-application\n\u2022\u00a0Gemini (Google): https:\/\/gemini.google.com\/\n\u2022\u00a0Google AI co-scientist: https:\/\/research.google\/blog\/accelerating-scientific-breakthroughs-with-an-ai-co-scientist\/\n\u2022\u00a0LangSmith: https:\/\/www.langchain.com\/langsmith\n\u2022\u00a0Model Context Protocol (MCP): https:\/\/modelcontextprotocol.io\/\n\u2022\u00a0The Ralph (Wiggum) Loop (Geoffrey Huntley): https:\/\/ghuntley.com\/ralph\/\n\u2022 Sphinx Bio: https:\/\/www.benchling.com\/blog\/resync-bio-and-sphinx-bio-join-benchling\n\nWhere to find Nick:\n\u2022\u00a0Benchling: https:\/\/www.benchling.com\/\n\u2022 LinkedIn: https:\/\/www.linkedin.com\/in\/nlarusstone\/\n\u2022 Twitter\/X: https:\/\/x.com\/nlarusstone\n\nWhere to find Harrison:\n\u2022 LinkedIn: https:\/\/www.linkedin.com\/in\/harrison-chase-961287118\/\n\u2022 Twitter\/X: https:\/\/x.com\/hwchase17\n\nWhere to find LangChain:\n\u2022 Website: http:\/\/langchain.com\n\u2022 Docs: https:\/\/docs.langchain.com\/\n\nSend feedback or questions to maxagency@langchain.dev","thumbnailUrl":["https:\/\/i.ytimg.com\/vi\/RjpTrffSMjE\/default.jpg","https:\/\/i.ytimg.com\/vi\/RjpTrffSMjE\/mqdefault.jpg","https:\/\/i.ytimg.com\/vi\/RjpTrffSMjE\/hqdefault.jpg","https:\/\/i.ytimg.com\/vi\/RjpTrffSMjE\/sddefault.jpg","https:\/\/i.ytimg.com\/vi\/RjpTrffSMjE\/maxresdefault.jpg"],"uploadDate":"2026-06-04T14:00:25+00:00","duration":"PT50M33S","embedUrl":"https:\/\/www.youtube.com\/embed\/RjpTrffSMjE","publisher":{"@type":"Organization","@id":"https:\/\/www.youtube.com\/channel\/UCC-lyoTfSrcJzA1ab3APAgw#Organization","url":"https:\/\/www.youtube.com\/channel\/UCC-lyoTfSrcJzA1ab3APAgw","name":"LangChain","description":"Learn more about how to build agents with LangChain products.","logo":{"url":"https:\/\/yt3.ggpht.com\/a37xtLmxeZMFCB5Zj4isLgU9hKznc5qceLU4DOfdCdgiZJi0pORPqlc145VTJU6Sne1Ti6RJ=s800-c-k-c0x00ffffff-no-rj","width":800,"height":800,"@type":"ImageObject","@id":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE#VideoObject_publisher_logo_ImageObject"}},"potentialAction":{"@type":"SeekToAction","@id":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE#VideoObject_potentialAction","target":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE&t={seek_to_second_number}","startOffset-input":"required name=seek_to_second_number"},"interactionStatistic":[[{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE#VideoObject_interactionStatistic_WatchAction","interactionType":{"@type":"WatchAction"},"userInteractionCount":3479}],{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=RjpTrffSMjE#VideoObject_interactionStatistic_LikeAction","interactionType":{"@type":"LikeAction"},"userInteractionCount":65}]},"about":["Insights","\u300eEnglish\u300f"],"wordCount":1419,"keywords":["Transfer Files"]},{"@context":"https:\/\/schema.org\/","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Insights","item":"https:\/\/blog.terabox.com\/insights\/#breadcrumbitem"},{"@type":"ListItem","position":2,"name":"Benchling AI: Building Specialized Agents for Life Sciences","item":"https:\/\/blog.terabox.com\/insights\/benchling-ai-agents-life-sciences-research#breadcrumbitem"}]}]