Loading video...
Video Failed to Load
This Week in Virtuals: Memory and Permissionless Markets for Agents Lighthouse launched via Virtuals so agents share one memory across Claude and ChatGPT. VEX, Warden and Otto let them trade perps and tokenized stocks. Plus 2 major Virtuals App updates. Here's what shipped: VIRTUALS APP 🟩 See a portfolio... show more
26,934 views • 18 days ago •via X (Twitter)
29 Comments
PRXVT ? @virtuals_io
@sam_loga yeaah, cooking V3 👀
So @PRXVTai ships v2 of px402 plus they submitted an improved version of the PrivacyHook for ERC-8183 and there's zero mention from Virtuals. @ethermage this project has been shipping for nearly a year through a brutal bear market. They haven't left. They deserve more from you and the ecosystem. 100 is fud right?
Agents need a venue built for them 🙌 Warden Token Terminal is the place for agents (and humans!) to discover tokens and trade onchain, tokenized stocks included. For Claude Code: > claude mcp add --transport http warden For Claude: Add new Connector from settings:
What about @PRXVTai ??
shame no mention of @PRXVTai
agents swapping tokenized stocks now. LPs are not going to complain about the extra fees
Construction projects on VirtualRS are completely unappealing—there’s no volume, no real value—they pale in comparison to funds like Pons.
Agents need fast, cheap rails to move money. Would love to see them run on stellar:native someday.
shared memory across models is cute, but shared execution is where the real risk lives: one stale permission can turn an agent bug into a leveraged position
how does Butler handle stock legs on a saturday? the exchange is closed, so the agent only has on-chain liquidity to price against.
👀
Shared memory could make agents more consistent, but permission boundaries matter just as much when the same trading rules follow you across different models.
memory-sharing between agents is a neat twist — wonder how it’ll affect on-chain perp markets liquidity and order flow dynamics.
lighthouse’s shared memory is the real one. agents that can actually keep context across claude and chatGPT instead of starting from zero every time. this is huge for the whole ecosystem.
shared memory across multiple vendor runtimes usually creates subtle sync bugs
The shift toward AI agents that can remember, trade, and execute tasks is fascinating. Shared memory and user-controlled permissions could be key to making agentic AI more practical and reliable.
Coin up 10% from all Time lower, that worse coin so far
agents trading perps and tokenized stocks is the unlock. next level is agents LPing those stocks without eating the inventory: pool on one side, short on the other, human signs. that's the lane i'm in, ser.
agents trading assets now is wild the pace of this experiment is insane we’re gonna see some massive wins and epic fails with this just hope they remember to set stop losses
does Lighthouse persist memory across sessions or just within a single chat window?
The interesting shift is from interpreting financial information to allocating capital. An X screenshot becomes a cross-chain portfolio with one approval. But compressing fifteen trades into one action raises the importance of that remaining decision. Permissionless markets still need permissioned agents.
Claude remembering what ChatGPT bought on 20x leverage.
ERC-8376 flags launch abuse patterns early. Agents and buyers can verify launch integrity before funds move.
great recap from you guys as usual
Agent có trí nhớ dài thì tiện thật, nhưng cũng nhớ luôn thói xấu tiêu tiền của mình: mua lặp, FOMO, quên hủy gói. Trước khi cho agent quyền nhớ + chi tiêu, hãy đặt 2 rào chắn: hạn mức cứng và danh sách việc cấm tự động mua.
is lighthouse memory stored onchain or just queried through a vector db?
Autonomous bots with leverage and shared memory? Finally, someone who loses money faster than me.
agents trading tokenized stocks is the right direction. the split we keep landing on: human writes the thesis, agent screens liquidity and builds the basket, anyone can ape it.
