reply @SheikhRaih100 That’s the goal, letting agents pick up right where they left off across different tools without needing a replay.
Quickstart is up if you want to test it out: https://walrus.xyz/memory
0·-Neutral
t
twitter9/7meme
quote: A hotel search yielding stale rates is bad state management. Exposing another user's prompt is a lack of access control.
If memory isn't encrypted, isolated, and verifiably updated at the storage layer, agents will keep leaking private context and serving stale outputs. https://x.com/chowtato/status/2096763154785054808 | Today Instinct spent 10 minutes looking up hotel options in NYC just to spit out stale data with prices that do not exist.
Recently it’s been slow, can’t complete forms, and seems to have exposed users’ private memories.
Trust in AI agents will make or break consumer adoption. Agents will need a memory environment that just works - encrypted, private, auditable. That’s @WalrusProtocol’s agentic storage and access control
0·-Neutral
t
twitter9/7meme
There could be 100 AI agents for every person within a few years, and maybe a million each within ten.
@AlliumLabs talks about what that actually looks like, and why those agents are going to need verifiable data underneath them to function.