Harvard seeks $100M for 'memory' AI lab
Harvard is launching a new AI lab that aims to build memory‑augmentation tools and is seeking $100 million in funding to pursue the work. The announcement surfaced on social platforms and sparked debate about long‑term memory augmentation research. (x.com)
A Harvard-linked startup led by neuroscientist Gabriel Kreiman is seeking about $100 million to build artificial intelligence tools aimed at helping people “remember everything.” (news.bloomberglaw.com) Bloomberg Law reported on April 10 that the company, Engramme, is in talks with investors and came out of stealth in March 2026. Kreiman previously taught at Harvard Medical School and has spent years studying memory, vision, and biologically inspired artificial intelligence. (news.bloomberglaw.com) (mbb.harvard.edu) The basic idea is not a chatbot that answers general questions. It is a personal memory system that stores details from your conversations, documents, photos, and location history so software can bring them back without a search box or prompt. (engramme.com) (iq.harvard.edu) Harvard’s own startup program described the earlier version of the project, Memorious, in September 2025 as a way to build a “Memorome,” or organized record of a user’s memories. The same article said founders Spandan Madan and Gabriel Kreiman interviewed more than 50 potential users, including older adults, project managers, and software developers, before building an iPhone beta app and an enterprise application programming interface. (iq.harvard.edu) That pitch lands in a wider race to give artificial intelligence systems longer memory. Most large language models still work like a conversation with a short attention span, so startups are trying to add durable records of past chats, files, and actions. (engramme.com) (hai.stanford.edu) Engramme says its product is meant to be “searchless” and “promptless,” meaning the system should surface relevant memories on its own instead of waiting for a user to ask. The company says it is building “novel AI models of human memory,” and a recent product post said its memory application programming interface is in beta for developers and teams. (engramme.com) (testingcatalog.com) The company’s privacy policy, updated March 16, says Engramme can collect personal data shared through its services and through “Connected Accounts,” alongside messages exchanged with the service. That matters because a product built to remember conversations, places, and files depends on collecting unusually intimate data in the first place. (engramme.com) (iq.harvard.edu) Supporters frame that as a cognitive aid for people who lose track of names, meetings, and documents. Critics on social platforms have framed the same idea as a privacy risk, especially if a memory assistant becomes always-on software that watches daily life closely. (iq.harvard.edu) (x.com) Kreiman’s lab at Harvard has long worked at the intersection of computational neuroscience and artificial intelligence, with research spanning memory and language. The fundraising push suggests investors are now being asked to back that academic work as a full-scale commercial bet on memory as the next layer of artificial intelligence. (mbb.harvard.edu) (news.bloomberglaw.com)
