--- title: Video Has No Memory. Here's How We Built One. — James... description: James Le from TwelveLabs presents a framework for building a memory layer for video AI systems. He argues that video is not a bag of frames but a... canonical: https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs | daily.dev og:description: James Le from TwelveLabs presents a framework for building a memory layer for video AI systems. He argues that video is not a bag of frames but a... og:url: https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9 og:image: https://api.daily.dev/og/posts/KLlim7lS9.png og:image:alt: Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 20 min read · 0 upvotes · 0 comments ## Summary James Le from TwelveLabs presents a framework for building a memory layer for video AI systems. He argues that video is not a bag of frames but a spatial-temporal volume requiring continuity across visual, audio, motion, and metadata modalities. Current AI stacks fail video because they use wrong context (forcing video into text tokens), wrong memory (text-style vector search lacks durable continuity), and weak native video reasoning. TwelveLabs addresses this with a stack including semantic chunking, a multimodal embedding encoder (Marengo), a spatial-temporal context store, and a video-context-aware language model (Pegasus). The talk introduces the concept of a 'context graph' — a durable, queryable structure linking moments, entities, appearances, relationships, and corpus-level themes. Five principles for video memory are outlined: ingest once and reason many times, store primitives not just answers, ground every claim to timestamps, let intent shape memory, and keep the layer composable. Live demos show the system applied to World Cup footage, traffic surveillance, and advertising analysis. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Memories.ai is building the visual memory layer for wearables and robotics](https://daily.dev/posts/memories-ai-is-building-the-visual-memory-layer-for-wearables-and-robotics-1txajaa6w) · TechCrunch · 0 upvotes · 0 comments - [Memory for AI Agents: A New Paradigm of Context Engineering](https://daily.dev/posts/memory-for-ai-agents-a-new-paradigm-of-context-engineering-ku2rvmlcw) · The New Stack · 2 upvotes · 0 comments - [Agent Memory: Why Your AI Has Amnesia and How to Fix It](https://daily.dev/posts/agent-memory-why-your-ai-has-amnesia-and-how-to-fix-it-nkpwfuhgz) · Tech Lead Digest · 1 upvotes · 0 comments - [The Agentic Blueprint: Why Memory \(Not Context\) is the Next Frontier of AI](https://daily.dev/posts/the-agentic-blueprint-why-memory-not-context-is-the-next-frontier-of-ai-kfzx6yhj8) · Medium · 2 upvotes · 0 comments --- Tags: [#vector-search](https://daily.dev/tags/vector-search) [View this post on daily.dev](https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs","url":"https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9"},"datePublished":"2026-07-23T02:22:17.551Z","dateModified":"2026-07-23T02:22:40.975Z","description":"James Le from TwelveLabs presents a framework for building a memory layer for video AI systems. He argues that video is not a bag of frames but a...","image":"https://i.ytimg.com/vi/mOf-PP4mVjA/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/mOf-PP4mVjA/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/video-has-no-memory-here-s-how-we-built-one-james-le-twelvelabs-kllim7ls9","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"vector-search","timeRequired":"PT20M","video":{"@type":"VideoObject","name":"Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs","description":"James Le from TwelveLabs presents a framework for building a memory layer for video AI systems. He argues that video is not a bag of frames but a...","thumbnailUrl":"https://i.ytimg.com/vi/mOf-PP4mVjA/sddefault.jpg","uploadDate":"2026-07-23T02:22:17.551Z","duration":"PT20M","url":"https://api.daily.dev/r/KLlim7lS9","embedUrl":"https://www.youtube.com/embed/mOf-PP4mVjA"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs"}]} ```