[{"@context":"https:\/\/schema.org\/","@type":"BlogPosting","@id":"https:\/\/blog.terabox.com\/insights\/continual-learning-ai-memory-ngram-weights#BlogPosting","mainEntityOfPage":"https:\/\/blog.terabox.com\/insights\/continual-learning-ai-memory-ngram-weights","headline":"Continual Learning: How Ngram Bakes Memory into AI Weights","name":"Continual Learning: How Ngram Bakes Memory into AI Weights","description":"\ud83d\udcfa Today’s recommended deep-dive video: https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY The Always-Training Model: Why Ngram is Building “Neural Memory” for the EnterpriseThe End of the “Smart Stranger”The Architecture of InternalizationLessons from the Human BrainKey TakeawaysQ&A The Always-Training Model: Why Ngram is Building “Neural Memory”... ","datePublished":"2026-07-23","dateModified":"2026-07-23","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\/aiR7F4jqjXY\/maxresdefault.jpg","url":"https:\/\/img.youtube.com\/vi\/aiR7F4jqjXY\/maxresdefault.jpg","height":"","width":""},"url":"https:\/\/blog.terabox.com\/insights\/continual-learning-ai-memory-ngram-weights","video":{"@context":"http:\/\/schema.org\/","@type":"VideoObject","@id":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY#VideoObject","contentUrl":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY","name":"Memory and Continual Learning: Engram's Dan Biderman and Jessy Lin","description":"Dan Biderman and Jessy Lin, co-founders of Engram, are building a neolab around memory and continual learning, which they call two sides of the same coin. Their contrarian premise: instead of stuffing ever-larger prompts into the context window or bolting on RAG, bake a team's knowledge directly into the model's weights, so it knows your company the way an employee of several years does. \n\nThe payoff: matching or beating frontier models while consuming up to 100x fewer tokens. Working with partners like Microsoft, Notion, and Harvey, the team draws on roots in computational neuroscience and state-space architectures to attack what they see as the real bottleneck in AI \u2014 not raw intelligence, but memory and continual learning. In contrast to the frontier labs' race toward one ever-bigger model and AGI, Dan and Jessy imagine a world where everyone has their own model \u2014 privately trained, always learning, and good at the things you actually care about. The real ChatGPT moment for memory, they argue, is the day your model feels like an intern that genuinely got smarter overnight.\n\n\nHosted by Sonya Huang and Shaun Maguire, Sequoia Capital\n\n00:00 Introduction\n00:59 Always Training Explained\n01:51 Beyond Context Windows\n03:29 Ngram Product Overview\n04:34 Adapters And Training Signals\n05:32 Internalize Vs Externalize\n06:49 Compute And Token Savings\n08:19 Teams First Then Individuals\n08:51 Memorization Vs Understanding\n12:47 Dreams And Offline Digestion\n14:08 Training Beats Curation\n15:19 Why Everyone Needs A Model\n21:44 Bitter Lesson And Architecture\n24:44 RAG Killer And KV Cache\n31:38 Future Of Memory And Models","thumbnailUrl":["https:\/\/i.ytimg.com\/vi\/aiR7F4jqjXY\/default.jpg","https:\/\/i.ytimg.com\/vi\/aiR7F4jqjXY\/mqdefault.jpg","https:\/\/i.ytimg.com\/vi\/aiR7F4jqjXY\/hqdefault.jpg","https:\/\/i.ytimg.com\/vi\/aiR7F4jqjXY\/sddefault.jpg","https:\/\/i.ytimg.com\/vi\/aiR7F4jqjXY\/maxresdefault.jpg"],"uploadDate":"2026-06-24T12:00:39+00:00","duration":"PT44M52S","embedUrl":"https:\/\/www.youtube.com\/embed\/aiR7F4jqjXY","publisher":{"@type":"Organization","@id":"https:\/\/www.youtube.com\/channel\/UCWrF0oN6unbXrWsTN7RctTw#Organization","url":"https:\/\/www.youtube.com\/channel\/UCWrF0oN6unbXrWsTN7RctTw","name":"Sequoia Capital","description":"Sequoia helps daring founders build legendary companies from idea to IPO and beyond. We aim to be the first true believers in tomorrow\u2019s most consequential companies. We partner with a few outliers each year and go all-in, providing them with the hands-on help required at every stage of the company building journey. Our expertise comes from nearly 50 years of working with legendary founders like Steve Jobs, Elon Musk, Larry Page, Jan Koum, Brian Chesky, Tony Xu, Lin Qiao, Eric Yuan, Christina Cacioppo, and Patrick Collison. In aggregate, Sequoia-backed companies account for more than 30% of NASDAQ's total value. The vast majority of the money we invest has been on behalf of nonprofits and schools like the Ford Foundation, Mayo Clinic and MIT, which means most of the returns we generate benefit these great causes.\n\n","logo":{"url":"https:\/\/yt3.ggpht.com\/uCN-D7KzMQY-Ti-xTsNAwilXVFFMYjEBRju_mXrR22HUYxJZjVZgP_SnamO9KbPo2XN-nE3O-A=s800-c-k-c0x00ffffff-no-rj","width":800,"height":800,"@type":"ImageObject","@id":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY#VideoObject_publisher_logo_ImageObject"}},"potentialAction":{"@type":"SeekToAction","@id":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY#VideoObject_potentialAction","target":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY&t={seek_to_second_number}","startOffset-input":"required name=seek_to_second_number"},"interactionStatistic":[[{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY#VideoObject_interactionStatistic_WatchAction","interactionType":{"@type":"WatchAction"},"userInteractionCount":25344}],{"@type":"InteractionCounter","@id":"https:\/\/www.youtube.com\/watch?v=aiR7F4jqjXY#VideoObject_interactionStatistic_LikeAction","interactionType":{"@type":"LikeAction"},"userInteractionCount":279}]},"about":["Insights","\u300eEnglish\u300f"],"wordCount":1444,"keywords":["permanent"]},{"@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":"Continual Learning: How Ngram Bakes Memory into AI Weights","item":"https:\/\/blog.terabox.com\/insights\/continual-learning-ai-memory-ngram-weights#breadcrumbitem"}]}]