AI Startup Letta Launches from UC Berkeley with $10M Seed Funding

Letta, one of UC Berkeley’s most anticipated AI startups, has officially come out of stealth mode, showcasing technology designed to help AI models remember users and conversations. The company announced $10 million in seed funding, led by Felicis’ Astasia Myers, at a $70 million post-money valuation. This launch is particularly noteworthy as Letta is a product of UC Berkeley’s Sky Computing Lab, a renowned incubator for groundbreaking AI projects.

Backed by Prominent Investors

Letta has garnered support from a prestigious lineup of angel investors in the AI space, including Google’s Jeff Dean, Hugging Face’s Clem Delangue, Runway’s Cristóbal Valenzuela, and Anyscale’s Robert Nishihara. The startup, founded by Berkeley PhD students Sarah Wooders and Charles Packer, has gained significant attention due to its origins in Berkeley’s Sky Computing Lab, the commercial home of the popular MemGPT open-source project.

Addressing AI Memory Limitations

MemGPT, a foundational project of Letta, addresses a key limitation in large language models (LLMs): the inability to retain long-term memory. Traditional models, like ChatGPT, are stateless and cannot remember past interactions, posing challenges for applications that require ongoing user engagement and personalization. MemGPT’s technology manages data and memory, enabling AI agents and chatbots to remember users and conversations over time. This capability is crucial for applications ranging from customer support to healthcare, where continuity of information is essential.

Viral Success and Investor Interest

MemGPT gained viral attention even before its official release. The founders had posted a whitepaper on Thursday, October 12, 2023, with plans to release a more detailed paper and the code on GitHub the following Monday. However, before they could make the official release, an anonymous user discovered the whitepaper and reposted it on Hacker News on Sunday, where it quickly went viral. The project attracted widespread interest from the AI community and investors alike, with the initial Hacker News post remaining at the top of the site for 48 hours.

“We hadn’t even had the chance to properly release the code, release the paper, or do a tweet thread," before it went viral, Packer recalled. The project’s GitHub page soon amassed over 11,000 stars and 1,200 forks, solidifying its reputation as a groundbreaking tool in AI development.

Felicis’ Astasia Myers, who led Letta’s seed funding round, discovered the project through its viral exposure and recognized its commercial potential. “We had an investment theme around AI agent infrastructure and appreciated that a really important component of that was the data and memory management to make these conversational chatbots and AI agents effective,” Myers said.

Future Plans and Open Source Commitment

Letta is positioning itself as a strong advocate for open-source AI development, offering an alternative to proprietary models like OpenAI’s. The startup’s commercial variant, Letta Cloud, is not yet available but is accepting beta user requests. Letta Cloud will offer a hosted agent service that allows developers to deploy and run stateful agents in the cloud, providing long-term memory storage and developer tools for building AI agents.

Competing in a Crowded Market

While Letta has made a significant impact with MemGPT, it faces competition from other AI agent-making tools, such as LangChain and OpenAI’s Assistants API. OpenAI’s latest model, the o1, is also addressing state management, potentially making memory solutions less critical for some users. Because it operates through multiple steps, the model inherently needs to maintain some level of state to process information and verify facts before generating a response. However, Letta’s open-source approach and compatibility with multiple AI models set it apart, as it aims to provide a more transparent and flexible alternative for developers.

“We are positioning ourselves as the open alternative to OpenAI,” said Packer. “I think it’s actually very, very hard to build very good AI applications, especially when you care about things like hallucination, if you can’t see what’s going on under the hood.”

A Promising Future

Letta’s emergence marks a significant development in the AI landscape. With strong backing, innovative technology, and a commitment to open source, the startup is poised to make a substantial impact. As Letta Cloud prepares for launch, it will be interesting to see how the company navigates the competitive landscape and further advances the capabilities of AI agents.