[{"@context":"https:\/\/schema.org\/","@type":"BlogPosting","@id":"https:\/\/blog.terabox.com\/insights\/defensa-cibernetica-autonoma-agentes-ia#BlogPosting","mainEntityOfPage":"https:\/\/blog.terabox.com\/insights\/defensa-cibernetica-autonoma-agentes-ia","headline":"Defensa Cibern\u00e9tica Aut\u00f3noma con Agentes de IA | Cogent","name":"Defensa Cibern\u00e9tica Aut\u00f3noma con Agentes de IA | Cogent","description":"\ud83d\udcfa V\u00eddeo de estudio recomendado hoy: https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g Defensa Cibern\u00e9tica Aut\u00f3noma: C\u00f3mo Cogent est\u00e1 Cerrando Vulnerabilidades a Velocidad de M\u00e1quinaLa Carrera contra el Tiempo: De Meses a MinutosLa Arquitectura del Contexto: Swarms y GrafosLa Escalera de Confianza y Remediaci\u00f3nEvaluaci\u00f3n y Memoria:…","datePublished":"2026-07-16","dateModified":"2026-07-16","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\/D6XWu54oG4g\/maxresdefault.jpg","url":"https:\/\/img.youtube.com\/vi\/D6XWu54oG4g\/maxresdefault.jpg","height":"","width":""},"url":"https:\/\/blog.terabox.com\/insights\/defensa-cibernetica-autonoma-agentes-ia","video":{"@context":"http:\/\/schema.org\/","@type":"VideoObject","@id":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g#VideoObject","contentUrl":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g","name":"Inside Cogent's three-agent architecture for autonomous defense | Geng Sng (Co-founder, Cogent)","description":"Geng Sng is co-founder and CTO of Cogent, which builds autonomous agents that remediate vulnerabilities for enterprise security teams. Today, Cogent's agents process billions of security events per day, maintaining a live context graph of every asset and vulnerability across customer environments. In this conversation, Geng walks through Cogent's hot vs cold context split, the sub-agents that handle side quests, and the two graphs they run in parallel.\n\nWe also discuss:\n\u2022 Why defensive security is harder for AI than offensive\n\u2022 Under the hood of Cogent's three agents\n\u2022 Inside Cogent's \u201cread only\u201d by-default sandboxes\n\u2022 Why graph databases don't scale for security data\n\u2022 Cogent Research and the move into formal verification\n\u2022 Why interactive agents need a deeper planning phase to one-shot\n\nReferences:\n\u2022 Abnormal AI: https:\/\/abnormal.ai\/\n\u2022 Amazon S3: https:\/\/aws.amazon.com\/s3\/\n\u2022 Anthropic: https:\/\/www.anthropic.com\/\n\u2022 Bash: https:\/\/www.gnu.org\/software\/bash\/\n\u2022 ChatGPT: https:\/\/chatgpt.com\/\n\u2022 Claude Code: https:\/\/www.anthropic.com\/claude-code\n\u2022 Claude Mythos: https:\/\/red.anthropic.com\/2026\/mythos-preview\/\n\u2022 CodeMender: https:\/\/deepmind.google\/blog\/introducing-codemender-an-ai-agent-for-code-security\/\n\u2022 Codex: https:\/\/openai.com\/codex\/\n\u2022 Cogent: https:\/\/www.cogent.com\/\n\u2022 Cursor: https:\/\/cursor.com\/\n\u2022 Google DeepMind: https:\/\/deepmind.google\/\n\u2022 GPT-5.5-Cyber: https:\/\/openai.com\/index\/gpt-5-5-with-trusted-access-for-cyber\/\n\u2022 Jupyter: https:\/\/jupyter.org\/\n\u2022 Letta: https:\/\/www.letta.com\/\n\u2022 Mozilla: https:\/\/www.mozilla.org\/\n\u2022 OpenAI: https:\/\/openai.com\/\n\u2022 Opus 4.6: https:\/\/www.anthropic.com\/news\/claude-opus-4-6\n\u2022 Opus 4.7: https:\/\/www.anthropic.com\/news\/claude-opus-4-7\n\u2022 Vercel: https:\/\/vercel.com\/\n\nWhere to find Geng:\n\u2022 LinkedIn: https:\/\/www.linkedin.com\/in\/geng-sng\/\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\n\nTimestamps:\n00:00 Why mean time to exploit collapsed from years to minutes\n02:08 Inside Cogent's Agent Lake architecture\n05:11 Why Cogent rejected graph databases\n10:48 The trust ladder before agents touch production\n15:13 The three types of agents inside Cogent\n17:07 How Cogent sandboxes its agents\n19:16 Short-circuiting interactive agents with a deeper planning phase\n24:31 What to do when users believe agents too much\n31:21 Why sub-agents let agents go on side quests\n34:59 Two-tiered evals and the metric that catches bad prompts\n40:00 Cogent's unique approach to context\n48:39 Cogent Research and the move into formal verification\n51:33 The single trait Cogent hires for\n54:00 Open-sourcing models within six months\n57:07 Why defensive security won't be commoditized anytime soon\n1:00:51 The founding insight behind Cogent","thumbnailUrl":["https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/default.jpg","https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/mqdefault.jpg","https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/hqdefault.jpg","https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/sddefault.jpg","https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/maxresdefault.jpg","https:\/\/i.ytimg.com\/vi\/D6XWu54oG4g\/fhddefault.jpg"],"uploadDate":"2026-05-22T14:00:13+00:00","duration":"PT1H14M58S","embedUrl":"https:\/\/www.youtube.com\/embed\/D6XWu54oG4g","publisher":{"@type":"Organization","@id":"https:\/\/www.youtube.com\/channel\/UCC-lyoTfSrcJzA1ab3APAgw#Organization","url":"https:\/\/www.youtube.com\/channel\/UCC-lyoTfSrcJzA1ab3APAgw","name":"LangChain","description":"LangChain enables every company to own their intelligence. LangChain's open, model-agnostic harnesses give teams choice and control over how they build their agent architecture. LangSmith brings testing, deployment, and monitoring together so teams can continuously improve their agents, compound intelligence, and govern them at scale. More than 7,000 customers, including Nvidia, Bridgewater, LinkedIn, Workday, Harvey, and Rippling trust LangSmith to improve their agents across the Agent Development Lifecycle.\n","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=D6XWu54oG4g#VideoObject_publisher_logo_ImageObject"}},"potentialAction":{"@type":"SeekToAction","@id":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g#VideoObject_potentialAction","target":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g&t={seek_to_second_number}","startOffset-input":"required name=seek_to_second_number"},"interactionStatistic":[[{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g#VideoObject_interactionStatistic_WatchAction","interactionType":{"@type":"WatchAction"},"userInteractionCount":5547}],{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=D6XWu54oG4g#VideoObject_interactionStatistic_LikeAction","interactionType":{"@type":"LikeAction"},"userInteractionCount":152}]},"about":["Insights","\u300eSpanish\u300f"],"wordCount":2090},{"@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":"Defensa Cibern\u00e9tica Aut\u00f3noma con Agentes de IA | Cogent","item":"https:\/\/blog.terabox.com\/insights\/defensa-cibernetica-autonoma-agentes-ia#breadcrumbitem"}]}]