Don’t prompt. Leave it to Prime.
Today we’re introducing Prime : CUA that autocompletes your actions. Press a hot key and go get some coffee. Come back and see what Prime has done for you.
Sign up for closed beta: discord.gg/fqwAmDxa9
Language models got smarter.
Humans got busier, not freer.
Agents work faster than people — but they keep stopping. To ask. To clarify. To wait for review.
The binding constraint stopped being model performance. It's human time and attention.
Today, we publicly release HEVEC, a vector database built on homomorphic encryption, enabling end-to-end privacy with real-time search at scale.
💻 Code: lnkd.in/gTzWC_Bz
📄 Paper: lnkd.in/g45fEaFD
What HEVEC enables:
• A secure, drop-in alternative to plaintext vector databases
• End-to-end homomorphic encryption for both data and queries
• Real-time encrypted search at scale (1M vectors in 187 ms)
As personal AI agents become deeply personalized, data ownership must belong to users.
HEVEC enforces this through privacy-by-design architecture.
🙌 We’d love feedback from the AI, systems, and privacy communities.