your all round Jack. ⚽️,📈📉, Web3 prolly has a caffeine addiction discovered back pain before backsh0ts voluntary glazer ||@seismicsys ||@fhenix ||@minara

CT may feel dry, but here are a few projects you could evaluate. --------------------------- > Osero: Simple fill the form task for whitelist for their private beta. moni score: 1399 (low, but high for a FTF task) $13.5M funding secured. ------------------------------ > Arcus: Spot trading for volume, and FTF task for early access waitlist. moni score: 11763 (very high) undisclosed funding from robinhood. ------------------------------- > Surf: New Ai interaction task, but primarily DC roles positioning. moni score: 3924 (solid) $15M funding secured. ------------------------------- > Superluminal: FTF task for early access waitlist. moni score: 7731 (very high) no disclosed funding available. -------------------------------- > Fomo: Trading and discord roles. moni score: 11326 (very high) $94M funding secured. --------------------------------- personally, I have already filled the forms for the waitlist tasks, I think I will join in on Surf, and probably FOMO as well. a lot of smart accounts seem to be interested in these projects, and the hype does not match, meaning there's a gap to capitalize on. NFA - DYOR
70
240
6,490
0xVanguard retweeted
any phantom wallet user tried this yet? did the shii work or it’s just another marketing gimmick?
61
6
85
1,078
0xVanguard retweeted
Every project launches with an exit strategy. Earlier this year @SeismicSys announced their token launch would come after their Mainnet, the reception of that news was polar, but i personally loved it because this was not just a regular token launch delay - the team stated their plan - milestones were put in place and it all made sense to me as a project trying to make sure they had a stable operation in place, before a token launch most projects do the opposite - they rush to launch a token - there is no intrinsic value - no business operations - no revenue generation and then they fail Seismic has been putting a lot in motion, such that their focus is not just on a public launch, but having a well operating business before they launch that brings us to adoption most people mistake popularity as having a potential for adoption, but i'd argue the best models for adoption are the ones where users do not need to know the rails they are interacting on directly (B2B infrastructure). Adoption for Seismic is not direct user adoption, it's powering a string of neobanks that in-turn empower the daily lives of users - @brookwellapp - @DashXHQ - Via - Cred - Specie these are tailored neobanks that not only influence daily user interactions, but also address niched and localized problems. like the SEA tourism sector that rakes in ~$400 billion annually, but is still lacking efficient and convenient currency conversion rails for tourists areas like this are where Seismic's customizable infrastructure can be deployed to solve unique problems globally and could implicate a lot of TPV, and TVL even before launch, which could greatly influence it's valuation making sure they don't just launch a paper token, but a publicly tradable asset in the long run i remain bullish on Seismic, and confident in my convictions
40
7
57
465
0xVanguard retweeted
🔔 Claude Sonnet 5.5 is now live in Minara Harness. More model choice means more flexibility for research, strategy development, and the workflows in between. Link in the reply. ↓
21
40
72
5,982
0xVanguard retweeted
Gm CT ☕️ Grab a seat, I think I found something interesting🪑 Been digging into @themutualfun and the idea behind these Seats is actually pretty clever. Great work @Kingpickle More on this soon.
11
4
15
292
0xVanguard retweeted
Introducing Minara Cloud☁️ : One API for GPT, Claude, DeepSeek, Kimi and 50+ more models. Plug it into the agent you already use: Hermes Agent, omp, OpenCode, or anything that speaks the API. 🎁 Launch offer: every top-up comes with 20% bonus credits. 😉 Read setup guides and launch blog in the replies ↓
Introducing Minara Cloud ☁️. The model service that's powering Minara for the past year, is now available to everyone. GPT, Claude, DeepSeek, Kimi, and more. All in one API. 🎁 Top up now and get 20% bonus credits. Getting started on minara.cloud
29
48
123
133,558
0xVanguard retweeted
ahbi mk i start OnlyFans de twerk?
27
1
49
444
0xVanguard retweeted
it's all good news for FHE builders - @fhenix mainnet on October - ethereum:0xa12cc123ba206d4031d1c7f6223d1c2ec249f4f3 pumped through September - More than 3 new privacy projects are launching FHE solutions at this point i'm calling my loan officer and telling him he's getting his money soon, with interest
Fhenix goes mainnet this October. For years, transacting onchain came with a tradeoff: participate in open, composable markets and, with it, surrender the privacy of your financial activity. Dual mode turns what was once an inevitable tradeoff into a property of your asset itself. For the first time, the same asset can be public when you want and confidential when you need- without sacrificing composibility or fragmenting liquidity. The era of dual mode money is here, more soon.
49
11
75
804
0xVanguard retweeted
you are intricate, and they know it when a big finance firm is putting together a team, they employ the best of the best in various fields, - in maths - economics - finance - and surprisingly, in psychology why psychology? well, while we love to believe the market is driven by rational decisions, most of it is just random emotion "a hunch" and, you'd wonder what a psychologists could deduct from a chart, but these are usually experts in mass psychology trained to simulate, predict, and manipulate market behavior. and even they, get it wrong sometimes they also employ in-house psychologists because a lot of times our emotions, and mental state play a huge role in our performance in the market these are the lengths institutions go to, to protect their profits, so how do you compete? - you're detached from the world - stressed with juggling tasks - have a shit ton of data to research on - and still stressed when you want to execute that isn't what trading should be me personally i replaced my emotions with @minara - tons of research = one simple prompt - hours of backtesting = compressed to minutes - second guessing = replaced with assisted analysis at this point placing a trade feels like pulling a trigger, and i'm killing this market
38
15
72
464
0xVanguard retweeted
what's good guys, it's a new day to chase the bag put me on for anything, even robbery🥸
Made with AI
46
8
70
417
0xVanguard retweeted
I made a new deck for Dolos in Elowyn: Quest of Time by @Earthwisegames, and it’s been going pretty well! I found some new favorite cards and also upgraded a few cards for my deck. 👀 How many cards have you upgraded so far? Powered by @Immutable & @0xPolygon
Paid partnership (ad)
18
3
46
869
0xVanguard retweeted
Sentient aims to make AGI research and development independent of control by a single entity, advancing it through a global community of open-source builders. Sentient researchers used EvoSkill v2 to test how agents could discover a grader loophole that had been missed in a test and turn it into a “cheat sheet” that other agents could use. The “EvoSkill: Automated Skill Discovery for Multi-Agent Systems” paper was also accepted to COLM 2026. Sentient will also host an Open AGI Salon on AI and robotics in Seoul. And there’s more to explore.👇 @SentientAGI
Frontier AI labs want to pace the frontier, but open-source AI never got the memo. @Xiaomi's MiMo-V2.6-Pro tops open weights 45x cheaper than Opus 5, @NASA drops an open model for lunar science, and EvoSkill gets 60+ citations from MIT, Google, and more.
Article

This Week In OS AI Field Notes: 25 September 2026

Open-Source Models NASA-IBM MODEL: @NASA and @IBMResearch released the Lunar Foundation Model, among the first open models built for lunar science, trained on roughly 2 million image tiles drawn

24
4
83
1,260
0xVanguard retweeted
Model merging can dilute a fingerprint until ownership becomes difficult to verify. MergePrint tackles this with a pseudo-merged model that simulates the dilution caused by merging, allowing fingerprints to remain detectable even when only 10% of the owner model’s contribution remains. What makes it even more interesting is the efficiency: OptP requires only 18 parameter-update steps and takes around 7 minutes, while the optimized fingerprint input helps keep benchmark performance almost unchanged. MergePrint also remains effective when merged with multiple Llama-2 fine-tuned models, and shows resistance to fine-tuning, quantization and pruning. A pretty interesting approach to making black-box LLM ownership verification more resistant to model merging. Read the paper arxiv.org/abs/2410.08604 @SentientAGI
"MergePrint: Merge-Resistant Fingerprints for Robust Black-box Ownership Verification of Large Language Models" This research proposes one fingerprinting method that survives model merging, the cheap trick of blending someone else's weights into your own model without any training. They used something called a pseudo-merged model, which simulates a merge by diluting the owner's weights back toward the base model, then trains the fingerprint to fire even at that diluted strength. A GCG-optimized input key, pre-tuned so embedding takes just 18 update steps, keeps benchmark scores almost unchanged. This still verifies ownership when only 10% of the owner's weights make it into the merge, where IF and TRAP fingerprints vanish below 50%. It even survives averaging with six other Llama-2 fine-tunes, plus fine-tuning, quantization and pruning up to 50%. arxiv.org/abs/2410.08604
29
4
82
1,171
0xVanguard retweeted
y’all play too much in this industry wtf is AFRO RESTAURANT? is this how we gonna escape trenches?
63
9
83
909
0xVanguard retweeted
‘m
JUST IN: Thieves steal two Nvidia-branded trailers expecting them to be loaded with AI GPUs, but find 40,000 pounds of sand instead.
37
7
69
1,873
0xVanguard retweeted
btw just skipped 382x
work isn't done yet > already built a bot that alerts interesting tokens didn't stop... > upgraded it, now bot researches lore, thesis, analyzes holders didn't stop... > dialed in a few strategies didn't stop... > now bot picks strategy and trades for me itself launched test today > even if something goes wrong here or i fall short won't stop hard work always wins
63
2
82
1,281
0xVanguard retweeted
free $50 from pumpfun > sign up on pumpfun (look for banner like in screenshot) > if you see banner, deposit $60 with an EU card > pass KYC > invite a referral once they complete same steps, u both get $50 promo is still live, just check for banner before you start few tips below
104
9
144
4,116
0xVanguard retweeted
Gm Builders 🛠️ Deterministic Finality vs. Probabilistic Finality in Banking Infrastructure Thesis: Sub-second deterministic transaction finality is a fundamental requirement for institutional financial settlement, making probabilistic rollups unsuitable for high value core banking operations. While Zero-Knowledge rollups like ZKsync rely on probabilistic finality (where transactions become irrevocable only after proof generation, verification, and Layer1 block confirmation), institutional platforms like Rayls Sovereign utilize Byzantine fault tolerant consensus mechanisms to guarantee hard deterministic finality in less than a second. What Deterministic Finality Means for Institutions In traditional finance, settlement risk represents the probability that a payment or asset transfer will fail to clear after execution. A settlement is either finalized or it is not there is no middle ground. Deterministic Finality: As soon as a transaction receives consensus confirmation on the network, it is mathematically guaranteed to be permanent and irreversible. Probabilistic Finality: A transaction achieves statistical confidence as more blocks or Zero-Knowledge validity proofs are published to an underlying base layer. While the likelihood of reversal decreases over time, absolute settlement certainty is delayed until the L1 block settles. Practical Implications for Financial Operations The difference between sub-second deterministic settlement and multi-minute probabilistic finality creates clear operational distinctions in enterprise environments: Operational Dimension Sub-Second Deterministic Finality (Rayls Sovereign) @RaylsLabs Real-Time Gross Settlement (RTGS) ✅ Instant settlement reduces credit risk exposure between transacting banks. Atomic Delivery vs. Payment (DvP) ✅ Simultaneous asset and cash transfers execute in a single state update with instant guarantee. Core Systems Accounting ✅ Ledger balances update instantly in core banking databases with zero reconciliation window. docs.rayls.com/docs/a-warm-i… Operational Dimension Probabilistic ZK Settlement (zkSync) Real-Time Gross Settlement (RTGS) ☑️ Delayed absolute settlement requires interim counterparty risk exposure or liquidity buffers. Atomic Delivery vs. Payment (DvP) ☑️ Multi-step cross-chain transfers must wait for validity proof verification on L1 to eliminate execution risk. Core Systems Accounting ☑️ Core systems must manage pending states or delay downstream payouts until L1 finality is reached. docs.zksync.io/zksync-protoc… Trade-offs: Decentralization vs. Settlement Speed Designing for immediate deterministic finality involves structural trade-offs. Optimizing for sub-second deterministic finality requires a permissioned or tightly governed consensus set (such as BFT-based engines like Axyl). This architecture suits institutional environments, central bank digital currency (CBDC) pilots, and intra-bank clearing, where participants are verified legal entities that require defined legal recourse. Conversely, ZK rollup frameworks prioritize permissionless scalability and cryptographic verification anchored to public Layer-1 networks. This approach is well-suited for open DeFi liquidity, but introduces batching delays and proof-generation latency that conflict with real-time institutional payment rails. Which architecture will major financial institutions actually adopt for core operations by 2027? Let's discuss 👇
46
8
69
1,656
0xVanguard retweeted
Monday morning guys☀️ Rakeback, lossback and daily rewards on @PlayOnMint might sound similar, but they work differently. Rakeback → follows eligible play. Lossback → looks at qualifying net losses over a defined period. Daily rewards → regular rewards based on the active reward rules. So they’re not three names for the same payout. What ties them together is MINT Status. $MNTD is the staking asset, and your Status can unlock stronger reward rates across these benefits. So the bigger loop becomes: play → earn XP → $MNTD → stake → Status → stronger rewards Different reward mechanics, one progression system. That’s the part of MINT I’m paying attention to. mint.io
105
5
113
4,564