Engramme

Engramme is launching a public beta of its Large Memory Models (LMMs), an AI architecture designed to give applications persistent, lifelong memory of what a user has seen, said, and done. The pitch to developers is an API that works in the background across tools like email, calendars, and video calls, surfacing relevant context automatically rather than waiting for a search query or prompt. The launch matters because memory has become the limiting factor for useful AI agents, and an LMM is positioned as a new architecture that captures what you saw, who you talked to, and where you were, and surfaces the right piece back to you without explicit retrieval. The company is a Harvard spinout led by co-founder and CEO Gabriel Kreiman, a neuroscientist whose lab has spent two decades studying how the brain consolidates experience, alongside co-founder Spandan Madan, who completed his computer science PhD at Harvard working in Kreiman's lab . Engramme emerged from stealth recently and is reportedly in early talks to raise around $100 million , which sets the stakes for this beta as the first public test of whether the LMM concept holds up outside the lab. For founders and operators, the practical read is that Engramme wants to be infrastructure rather than a consumer app, sitting beneath other products and handling recall the way vector databases handle embeddings today. Beta access is open through the company website, and the early use cases shown in the launch (pulling up a saved restaurant when a friend invites you to lunch, surfacing prior project context inside an email thread, retrieving an answer mid-call) point at where the team expects the first integrations to land.