TractionFirst retweeted
robinhood:0xf44702b17d9abd53815f703e772f35e9c71a53af is the most complete AI + privacy project being built on Robinhood Chain.
Raxol: AI agents with their own wallets that can make payments and trade, but only within spending rules approved by a person.
Xochi: A private trading platform designed to keep your order out of the public queue, reducing opportunities for bots to see it and trade ahead of you.
Riddler: The execution engine that finds available liquidity across trading venues and completes the swaps requested through Xochi.
ERC-8262 / ZK Compliance: Their draft standard is designed to let applications verify identity-check credentials and financial screening requirements without exposing the sensitive information behind them.
Xochi SDK: A developer toolkit that lets other applications generate these privacy-preserving proofs without building the technology from scratch.
PXE Bridge: The connection to Aztec for deeper transaction privacy, with shielded trading through the bridge still being developed.
Mana: Their alternative Ethereum client, adding another implementation of the software that processes and verifies Ethereum transactions.
Node operations: The team also runs validators, blockchain data services and trading infrastructure across 19 networks.
The clever part is how the pieces fit together: Raxol brings the orders, Xochi and Riddler execute them, and the planned compliance layer checks specific requirements without making privacy disappear. Axol
And robinhood:0xf44702b17d9abd53815f703e772f35e9c71a53af has a defined place in that design: the team’s economic model connects orders routed by Raxol -> routing fees -> robinhood:0xf44702b17d9abd53815f703e772f35e9c71a53af buybacks
You should not expect less from the team that combines Lido and Blockdaemon protocol experience, General Dynamics engineering, and FBI/DOJ financial-intelligence and anti-money-laundering work.
Perfectly connected design where AI, privacy, and compliance serve the same transaction, with a token model tied to paid usage.
6
18
75
3,089
TractionFirst retweeted
underestimated the amount of un-fun work i had to do to prepare the product site 😅
still building out the back end for accounts and proper login systems for users to manage their devices and information.
also doing some full runs of LongMemEval_S and gathering additional data for our new benchmark design. we will be opening that public repo and doing official announcement sometime in mid to late October as well.
SIBYL mode
the new @sibyl_labs_ memory product page drops tomorrow.
what is coming with it:
card checkout. one tier, monthly or annual billing options. the first time the plugin can be bought with debit or credit card.
a straight funnel. try free, hit the cap, go pro. nothing deleted if you stop.
improved account management features and dashboard to manage your subscription and sessions.
this will mark lab's first dollar revenue rail from products. and the memory is the first product of many.
9
21
66
1,979
TractionFirst retweeted
A lot of big things on the horizon. This why my sleep schedule is non existent haha :)
3
9
27
832
$SIBYL 👀
1
27
499
I wouldn't hold $CPHY for much longer.
The market is moving on while the founder is crashing out. Every founder that crashes out on X nowadays kills the project by default. If you haven't learnt that to this day then there is no way saving you from what is about to happen.
6
109
Alpha for $NOCK / $EDGE / $TIBBIR hodlers
Ticker: SIBYL (on Virtuals)
Ca:
0x797f214a2CD64a4963A91Fa21c8C55Ec3EBa4714
we are beginning work on an open source memory benchmark test at @sibyl_labs_
after months of research we have concluded that the primary gauges the field uses for memory do not necessarily test the actual retrieval itself.
agentic environments play a large role in determining how models interact with "memory" or recalling facts.
a users agent.md could simply say "gregs hat is blue" which would remove all need for retrieval from the memory.
abstention rules are important for agents, as it teaches them that a non-answer is better than a confident hallucination.
but this has no affect on memory, or quality of the retrieval.
it is simply the agentic framework doing it's job, in combination with the memory.
so we have multiple moving pieces in an agents environment:
-persona/identity
-operating rules
-memory/retrieval
the way an agent presents the data from the retrieval process should be of no concern to people who want to gauge the performance of the memory. they are separate things.
this means the majority of "memory" benchmarks are in truth, framework benchmarks, or agentic operating systems benchmarks.
if you would like to contribute to this project, and help push the agentic memory, and agentic framework fields into the right direction with us, please feel free to reach out.
our only goal is to give the space an unbiased, objective benchmark format that allows memory companies to be properly scored and judged for metrics grounded in reality, instead of agent.md files.
we hope you want the same✊
7
594
I have terrible news for the $CPHY founder and his gullible followers accepting his deceitful behavior
Not sure why he is after Tulips but Tulips didn't say anything that wasn't untrue
Michael Joseph DID NOT invent the concept timechains
5
1
9
474
heard this one before from a dozen egocentric farmer devs that are all gone and forgotten.
YOU made $CPHY about YOU and thereby it WILL fail. why is a token needed? it's not needed. seems like you're a scammer just like the rest of them.
you have the physiognomy of a scammer.
5
183
robinhood:0xf44702b17d9abd53815f703e772f35e9c71a53af target in late Q4 = 500m mcap
Add it to your watchlist, $NEAR & $ZEC holders
Trending on @X: ZKPs in KYC/Compliance, thanks @HesterPeirce!
ERC-8262 is our answer, proposed as an open @ethereum standard: a Zero-Knowledge Compliance Oracle.
ERC-8262 computes a risk score from screening signals, then proves it sits below a jurisdiction's threshold.
The chain learns one thing: pass or fail. Scores, counterparties and amounts never leave your device.
If you'd like to learn more, we shared a detailed breakdown of how it works, why it's necessary, and what's next, pinned on our profile.
nitter.net/i/trending/21031619415…
4
27
3,147
$raxol builds agentic dark pools; privacy for ai agents. the two biggest metas united. as if Zec and Near had a baby
deployed on Robinhood, supported by eth core
just saying.
16
641
The $RAXOL team is lining up things to take the protocol to the next level.
This is a screenshot from a team post in the official Telegram group just minutes ago. The post is longer with even more alpha but I will not share it here on X.
EF = Ethereum Foundation 💡
2
11
542
we are beginning work on an open source memory benchmark test at @sibyl_labs_
after months of research we have concluded that the primary gauges the field uses for memory do not necessarily test the actual retrieval itself.
agentic environments play a large role in determining how models interact with "memory" or recalling facts.
a users agent.md could simply say "gregs hat is blue" which would remove all need for retrieval from the memory.
abstention rules are important for agents, as it teaches them that a non-answer is better than a confident hallucination.
but this has no affect on memory, or quality of the retrieval.
it is simply the agentic framework doing it's job, in combination with the memory.
so we have multiple moving pieces in an agents environment:
-persona/identity
-operating rules
-memory/retrieval
the way an agent presents the data from the retrieval process should be of no concern to people who want to gauge the performance of the memory. they are separate things.
this means the majority of "memory" benchmarks are in truth, framework benchmarks, or agentic operating systems benchmarks.
if you would like to contribute to this project, and help push the agentic memory, and agentic framework fields into the right direction with us, please feel free to reach out.
our only goal is to give the space an unbiased, objective benchmark format that allows memory companies to be properly scored and judged for metrics grounded in reality, instead of agent.md files.
we hope you want the same✊
1
6
522