Image Credits:Supermemory <\/p>\n
Context windows of AI models, which indicate the ability of a model to \u201cremember\u201d information, have increased over time. However, researchers have suggested<\/a> new ways<\/a> to increase long-term memory of AI models, as they often can\u2019t hold context over several sessions.<\/p>\n 19-year-old founder Dhravya Shah is attempting to solve problems in this area by building a memory solution, called Supermemory<\/a>, for AI apps.<\/p>\n Originally from Mumbai in India, Shah started building consumer-facing bots and apps a few years ago. He even sold his bot that formatted tweets into good-looking screenshots to the social media tool Hypefury.<\/p>\n The founder, who was preparing for an entrance exam to get into IIT(Indian Institute of Technology), earned good money from this sale and decided to move to the U.S to attend Arizona State University instead.<\/p>\n After relocating, he challenged himself to build something new each week for 40 weeks. During one of those weeks, he built Supermemory (which was initially called Any Context) and put it on GitHub. At that time, the tool allowed you to chat with your Twitter bookmarks.<\/p>\n The current version of the tool extracts \u201cmemories\u201d or insights from unstructured data and helps the applications understand the context better.<\/p>\n Shah secured an internship at Cloudflare in 2024, where he worked on AI and infrastructure. He later worked as a developer relations lead at the company. During this time, advisors, including Cloudflare CTO Dane Knecht, asked him to turn Supermemory into a product.<\/p>\n This year, he decided to build Supermemory full-time.<\/p>\n Supermemory, now described as a universal memory API for AI apps, builds a knowledge graph based on the data it processes and personalizes the context for the users. For instance, it can support querying across month-old entries for a writing or a journaling app, or searching for an email app. As the solution allows for multimodal inputs, it could also allow a video editor to fetch relevant assets from a library for a particular prompt.<\/p>\n The startup can ingest any type of data, the company says, including files, docs, chats, projects, emails, PDFs, and app data streams. Its chatbot and notetaker feature lets users add memories in text, add files or links, and connect to apps like Google Drive, OneDrive, or Notion. There\u2019s also a Chrome extension that lets you easily add notes from a website.<\/p>\n \u201cOur core strength is to extract insights from any kind of unstructured data and give the apps more context about users. As we work across multimodal data, our solution is suitable for all kinds of AI apps ranging from email clients to video editors,\u201d Shah said.<\/p>\n Supermemory has secured seed funding of $2.6 million led by Susa Ventures, Browder Capital, and SF1.vc. The round also includes individual investors like Cloudflare\u2019s Knecht, Google AI chief Jeff Dean, Deepmind product manager Logan Kilpatrick, Sentry founder David Cramer, and executives from OpenAI, Meta, and Google.<\/p>\n Shah said at one point, Y-Combinator also approached him to join one of its batches, but he already had investors on board, so the timing didn\u2019t work out.<\/p>\n Joshua Browder, founder and CEO of \u201cRobot lawyer\u201d startup DoNotPay, who runs Browder Capital as a solo GP, was impressed by Shah\u2019s tenacity.<\/p>\n \u201cI connected with Dhravya over X, and what struck me was how quickly he moves and builds things, and that prompted me to invest in him,\u201d he said.<\/p>\n The company has multiple existing customers, including a16z-backed desktop assistant Cluely<\/a>, AI video editor Montra<\/a>, AI search Scira<\/a>, Composio\u2019s multi-MCP tool Rube<\/a>, and Real estate startup Rets<\/a>. Plus, it\u2019s working with a robotics company to retain visual memories captured by a robot.<\/p>\n While this is a slant towards consumers, the app feels more like a playground for developers to understand more about the tool and potentially use it in their workflows or their own apps.<\/p>\n Supermemory has substantial competitors in the memory space. Startups like Felicis Ventures-backed Letta<\/a> and Mem0 (where Shah worked for a short while) are building a memory layer for agents. Supermemory\u2019s own backer, Susa Ventures, has invested in Memories.ai along with Samsung<\/a>, which can tap into thousands of hours of footage to get insights. Shah says that while these startups might serve different industries and use cases, Supermemory will stand out because it offers lower latency.<\/p>\n \u201cMore and more AI companies will need a memory layer. Supermemory\u2019s solution provides high performance while allowing you to surface relevant context quickly.\u201d Browder said.<\/p>\n","protected":false},"excerpt":{"rendered":"Image Credits:Supermemory Context windows of AI models, which indicate the ability of a model to \u201cremember\u201d information, have…\n","protected":false},"author":2,"featured_media":63745,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[220,218,219,7822,61,60,43735,43736,80,43734],"class_list":{"0":"post-63744","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-google-ai","12":"tag-ie","13":"tag-ireland","14":"tag-shah","15":"tag-supermemory","16":"tag-technology","17":"tag-unstructured-data"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/63744","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=63744"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/63744\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/63745"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=63744"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=63744"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=63744"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}} A memory map extracted by Supermemory Image Credits: Supermemory <\/p>\n
Image Credits: Supermemory <\/p>\n