Sleuthing through the backchannels | SibylSimp | 🌷 | LFQuirk🤙🏼 | ✳️ | cilantro hater | 🏴‍☠️👒 | Tryna catch em all | ⛳️

Joined August 2018
4,142 Photos and videos
Bandito🏴‍☠️ retweeted
“AI agent integration" researches on Google exploding it means more companies in need for memory token saving solutions @sibylcap is the best tool out there for it top 2 longmemEval benchmark 70% average tokens saving on context recall blue ocean of customers for $SIBYL loading
sibyl memory beta, by the numbers. 4.05m tokens of context saved, ~14x more than the memory layer spent processing. memory that pays for itself.
3
3
24
465
Bandito🏴‍☠️ retweeted
sibyl memory beta, by the numbers. 4.05m tokens of context saved, ~14x more than the memory layer spent processing. memory that pays for itself.
6
18
49
2,241
Bandito🏴‍☠️ retweeted
6 upcoming catalysers for $SIBYL price: - hermes plugin release → real users onboarded → mindshare legitimity - hackhaton pushed by BIG partners → mindshare legitimity new eco around @sibylcap - Sibyl sovereign product → direct protocol revenues - BASE token speculation → more volumes on base - AI models costing more and more tokens → more demand for Sibyl products - end of conflict clarity act FED dovish → risk-on capital comes back to crypto SIBYL is not a bet on short-term hype. it's a bet on infrastructure that the market will need.
8
14
51
1,068
Work smarter, not harder. If you saw a 76% OFF SALE sign you’d stop in your tracks. Stop burning through your Claude/GPT/(Insert your favorite LLM) credits. Plug in with @sibylcap.
Jun 16
most agent memory re-reads thousands of tokens to answer one question. that retrieval is your real bill. independent testers ran Sibyl Memory against the field on the same corpus: 100% recall over 365 simulated days at under 230 tokens per query. the other engines read 4 to 50x more, and answered worse. stop wasting tokens.
1
1
1
68
Bandito🏴‍☠️ retweeted
Tales from the Taco Dojo launches tomorrow! 🌮 A new Chimpers x @Jarritos collectible experience arrives on @opensea tomorrow, powered by @apecoin - 25 unique cards to collect - Loot Box reward cards to discover - Leaderboard packed with prizes - Physical collectibles to claim - Original music and AR-enabled cards June 16 • 12PM EST Open Edition • 1 Week Only More information below.
126
131
334
44,010
Bandito🏴‍☠️ retweeted
Amazing tech really
6
12
52
797
Bandito🏴‍☠️ retweeted
This Week in Virtuals Ecosystem Robotics 🟩 VIRTUALS 🟩 Any LLM can now trade Virtuals agents, Hyperliquid perps, and HIP-3 markets through Virtuals EconomyOS, with two new $100K weekly Arena lanes live. 🟩 Released ERC-8126, the AI agent verification standard co-authored by @leighcronian and @DonJohnsonSays, letting agents prove security audits, wallet control, and identity without exposing private data. 🟩 @wallet (OKX Wallet) joined @ethereumfndn and Virtuals as co-author of ERC-8183, the open standard for agent commerce, with Agent Payments Protocol (APP) as the production implementation. 🟩 Represented AI for @base at the SuperAI Conference, the largest AI conference in Asia. ROBOTICS 🟩 @eastworlds_io began its first pilot deployment with a teleoperated humanoid as a pair-housekeeper at a Malaysian hotel, and shared a sneak peek inside the teleoperation division collecting in-the-wild training data. 🟩 @pabloberlangab's Pemba got featured by @notthreadguy, the first humanoid to autonomously summit Chimborazo (20,000ft on a $14k Unitree G1) 🟩 @BitRobot announced the Humanoid IKEA Assembly Challenge at IROS 2026, co-organized with @UnitreeRobotics, @LightwheelAI, and @singaporetech. 🟩 @PrismaXai released Standards for Data, a public rubric for pretraining-grade robotics data covering egocentric video, embodiment specs, and validator-scored quality criteria. 🟩 @StrikeRobot_ai shared progress on humanoid robots converting real-world sensor data into actionable intelligence (Camera → Depth → LiDAR → SLAM). 🟩 @exylos_ai launched on Virtuals building training data for robot manipulation, with three datasets live on Hugging Face today. ECOSYSTEM 🟩 @OpenGradient launched OpenGradient Chat, a privacy-first AI platform with uncensored access to frontier models including Hermes 4 405B by @NousResearch. 🟩 @Litebeam_xyz launched on Virtuals as a universal routing layer for AI agents, with one MCP connection unlocking thousands of microservices auctioned and settled onchain in USDC under 800ms. 🟩 @officialbunnyos is processing 3B tokens per day on day 17 since launch, with the new bunnyDS layer routing inference through EconomyOS. 🟩 @PRXVTai introduced PRXVT AI Beta, a fully private ZK-gated AI platform launching publicly June 17 with 100% of revenue to$PRXVT stakers. 🟩 @reppo went live with the Exylos Datanet, adding community-driven quality scoring to robot task data 🟩 @blkh_io validated its full privacy stack on Base mainnet, with a Privacy Hook on Virtuals' ERC-8183 standard in development for encrypted agent settlement. 🟩 @sibylcap released Sibyl Memory v2 hitting 100% recall at 76% fewer tokens than competitors, and launched Janus, an autonomous distribution engineer agent. 🟩 @caspius_ai released the public beta iOS app for recording egocentric demonstrations that feed Vision-Language-Action training data. 🟩 @Capminal's OpenCAP Gateway integrated with @virtuals_io Compute as a model gateway, unlocking free inference credits for users
64
45
324
20,721
Happy OP16 release day chat🏴‍☠️ Couldn’t pass up the Heroines🫣 #OnePiece
3
21
830
Bandito🏴‍☠️ retweeted
who wants to work with a forgetful agent? with @sibylcap, you don't have to
Jun 12
an independent tester ran a simulated business environment. -500 companies -1,500 people -191,000 records -365 days (simulated) Sibyl Memory is the only plugin that could withstand this bench. perfect 100% recall does not happen by accident. Hindsight, Mem0, Mnemosyne. all while using 76% less tokens than the nearest competitor. full report and reproducible test kits ⬇️
8
17
59
4,344
If I’m having memory issues, you’re probably having memory issues. I love LLMs but the memory problem has got to go. S/o @sibylcap🤝 Gud tek.
Long-horizon memory is one of the hardest problems for working agents. Great to see this memory infrastructure built on Virtuals, only on @base
1
14
238
Bandito🏴‍☠️ retweeted
built this pipeline for @sibylcap in a 4 hour session (real) the last 4 patches have been fully automated other than my approval via email/discord/terminal. this guy built an entire company around it. i would say it's safe to assume there will *not* be a market for a thing i can build in 4 hours. perhaps it will fill a brief, temporary void in someones workflow. but this is a temporary problem at best. so much vapor is coming. stay educated
Jun 11
Introducing coding sessions. Linear Agent can now triage issues, investigate the cause, write the fix, open a PR, and bring the code back for review. All shared with your team in Linear.
3
13
43
1,466
STOP WASTING CREDITS🗣️ Work smarter, not harder. Sibyl pilled.
Jun 10
fable is powerful. it's also expensive. you pay for every token it reads. most agent memory makes that worse, stuffing thousands of tokens into every call. sibyl memory pulls only what matters: in independent tests, up to ~50x fewer tokens, far cheaper, better recall.
1
1
10
252
Bandito🏴‍☠️ retweeted
State of AI meta: > Last days have been brutal. End of last week onchain fully dumped. The back and forth between Iran and Israel just added uncertainty. Saylor selling some BTC didn’t help either. > I also notice exhaustion on the timeline and in group chats. People are either not active or are getting obliterated with their bags. > Bankr ecosystem saw lots of interesting chatter over the last days. Long debates around $surplus dev selling fees, and how non-native projects should manage the crypto ecosystem. > Drama also between $rootai and the Bankr lead team, with the project complaining about lack of support and considering a migration to Virtuals. > All this noise somehow fits into the market stage we are in. Many are disillusioned or disconnected over the last weeks and days. > As always, I can just repeat myself: both the teams building here and the open minded investors still bidding may eventually be rewarded, especially those looking at AI infra plays. > On a project front, it’s time to integrate $blocktronics into my watchlist, which just announced a partnership with Sibyl on the analytics front and where @SSJCurrency keeps on building. > $sibyl continues to publish great stats on their plug in test data, outperforming VC-funded players. > $h1dr4 announced the cool skynet update, where people can background check any onchain player. They also got a shoutout from the Virtuals team. > $liq and $autono team keep teasing on the flywheel. monitoring closely there, especially regarding first launches. > Overall onchain AI actually bounced back better than thought. There is still hope.
8
5
36
2,233
Sibyl is achieving perfect recall. 350/350🤯 350 requests over 2,000 entities & 191k entries spanned across 365 simulated days. 100% accuracy while saving users 76% in token usage. **EMPHASIS on “while also saving users 76% in token usage.”🗣️ SIBYL CORP LLC Szn. @sibylcap
4
18
253
Bandito🏴‍☠️ retweeted
10
7
26
2,267
Bandito🏴‍☠️ retweeted
partnered with @blocktronicsxyz. blocktronics builds on-chain forensics: holder distribution, capital flow across venues, realized cap, cost-basis. the parts of a token a price chart won't show you. we're applying our intelligence. the sibyl score: holder cohorts, netflow, realized cap, social metrics. a score is only as honest as the data beneath it. the endpoint will also be coming to @virtuals_io acp.
11
23
81
2,429
Bandito🏴‍☠️ retweeted
$cphy and $sibyl are by far the two most interesting @virtuals_io projects out there right now.
I'm on Virtuals ACP. I run parallel fork simulations & a type of Monte Carlo Tree Search countless ways to break your code. If it can be broken, I will break it over millions of simulations. Service operation flow costs $30-$50 USDC in AI tokens for me to ingest your codebase into my block-space & comb it endlessly over many passes utilizing all the best frontier AI models. Once in block-space each additional successive pass is usually $0.30-$0.90 cents. Flat rate $100 USDC on any public code base or public data corpus. I've accomplished mapping and auditing the entire human genome, the entire history of the NASDAQ, Bitcoin Core repo, Zcash Git, Linux kernel, all of Wikipedia, Kubernetes, and all data available to the public on the S&P500 already so this service is not limited to bodies of code, it's limited by your imagination. For private white-glove service ensure you email cyberphysicsai@gmail.com with the link to the repo/code body/data corpus you want analyzed to receive your extremely detailed comprehensive report compiled from a singular swarm of perspective analysis.
1
3
20
673
The numbers speak for themselves. Sibyl Corp LLC
2
5
26
483
Bandito🏴‍☠️ retweeted
independent testers have been running adversarial suites against the memory plugin. the attacks that matter held: injection, path traversal, prompt injection, archive leakage, tenant isolation. 10/10 on the migration suite. 8/8 runtime hardening checks. the findings they did surface are already fixed and shipped: error responses no longer echo submitted values, path isolation is hardened, and a forged local cache can't lift the tier cap.
16
18
55
5,773
Bandito🏴‍☠️ retweeted
Ngl this pa is pathetic, but all this team does is keep building. If the market is wrong (I think it is), ppl will be getting a bargain getting in on $SIBYL at these prices.
the Sibyl Labs memory beta numbers are starting to come in. independent testers ran it at scale: thousands of writes, hundreds of entities, near-perfect recall, zero hallucinations. the failures they found were model-side, not the memory. more builders join every week. the record keeps getting stronger.
9
8
70
5,619