san-francisco-ai
(4)
The Internet Priced Everything Except the Page
Zzazz wants every article, video and dataset to carry a live price. The strange part is not the number - it is the market the company hopes will form around it.
Charles Packer
Charles Packer is the co-founder and CEO of Letta, the San Francisco startup building the memory layer for AI agents. He is the original author of MemGPT, the viral research project that reframed large language models as operating systems that manage their own memory. A UC Berkeley computer science PhD out of the BAIR and Sky Computing labs, Packer walked away from offers at OpenAI and Google DeepMind to build stateful agents that remember, learn, and improve after deployment - the layer he argues sits above the base models and separates a real agent from a chatbot with a nice interface.
Personal AI
Personal AI (legally Human AI Labs, Inc.) builds an identity-based memory layer and Personal Language Models (PLMs) - compact, individually trained AI models that capture a person's or organization's knowledge, voice, and decision-making. Instead of a single generic chatbot, the platform lets people and enterprises create AI personas trained on their own data, keeping ownership and privacy with the user. Founded in 2020 and based in San Francisco, the company positions its smaller, memory-grounded models as a more accurate, lower-cost, privacy-first alternative to centralized cloud LLMs for sales, healthcare, legal, HR and knowledge work.
Zyphra
Zyphra is a superintelligence research and product company building efficient, open foundation models and a full-stack AI cloud. Best known for its SSM-hybrid Zamba models, the ZAYA1 reasoning family trained entirely on AMD hardware, and the Zonos text-to-speech system, Zyphra argues that frontier intelligence should run cheaply, locally, and outside the walled gardens of a handful of labs. In June 2025 it raised a $100M Series A at a $1B valuation led by Jaan Tallinn.