Shadow Zebra (@shadow_zebra) header
Shadow Zebra (@shadow_zebra) profile picture

The Zebra is an invisible horse with BOTH black and white stripes, so you don’t bump into him.

Solana
Joined October 2024
Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)Photo by Shadow Zebra (@shadow_zebra)
Perception is reality. Identity is a construct. Destiny is malleable. The individual is a useful illusion. Money doesn’t exist. Past and future aren’t real. Materialism is a religious cult. Spiritualism is relation to true material - rock/earth/fire/air.
🤖 Made with AI
1
3
1,550
Another day another 4k 🤷🏻‍♂️. I seen worse. Not disregarding $patchbay yet. dyor ofc HRa8h3NnN2tHLW3ndcnsrcATtMGH2i6Q54xiPaTcHbay
Three wallets have been paid more than 2 SOL between them for holding one coin. They bought the Patchbay coin and kept it. The rule did the rest. Every wallet the coins have paid is ranked here. patchbay.wtf/earners
53
👀👀 $vibestack 7YDiT1pxbazfzx6rkaUSSMEptJEwcWV2G8x9puSxpump
1
152
👀🤔 $dprf 7pZ5HMPTFQsLP6D8Zg6VfgjbKRycitG3H4Dq7cmBpump
Darkproof vs Telegram. Look beyond the lock icon. Telegram’s regular chats and groups aren’t end-to-end encrypted. Secret Chats are. Darkproof brings: ▪ End-to-end encrypted private groups ▪ No account or phone number required ▪ Built-in Tor, enabled by default, to hide your IP from relays ▪ Group keys that rotate when members join or leave ▪ On-device free AI, no cloud processing needed ▪ Offline messaging without a central server Your conversations deserve more privacy. → darkproof.app
1
264
$patchbay been cookin HRa8h3NnN2tHLW3ndcnsrcATtMGH2i6Q54xiPaTcHbay
Every coin page on Patchbay now has a Share on X button. One click opens a ready post with the coin's name, ticker and link. X adds the card with its rule. "Your share grows the longer you stay." Now that travels with the link. patchbay.wtf
123
🫡🔥🔥 F82MdJHkpFYW9kbokK9wq2xXYnB4BTkCYaWEi3WTpump
Good morning! Time to start a new day in $VICE 🌴 The first steps into Web3 are done: you can now log in and save your game with your wallet (Phantom or MetaMask on Solana). You can still play as a guest, but you may lose your game progress. Logging in with your wallet keeps your progress saved. #Solana
1
1
3
130
👀 $vice F82MdJHkpFYW9kbokK9wq2xXYnB4BTkCYaWEi3WTpump
Working on bringing $VICE tokenomics into the game itself ViceGame gets new updates every day for the coming weeks. Battle Royale just dropped in 1.8.5, and there's a lot more on the way. It's still early. Play now: vicegame.fun #solana #gta6
2
147
$dogcoin be like that best fling of your life walking away without saying goodbye like.. Are we not talking about it? I feel like we need to talk about it 🤣 🤷🏻‍♂️ rip.
66
$pdao be like.. 👀 Looks good sitting in that folder there eh
1
88
Mmk so once ppl are done fucking with these charts, idc which, the $makenzi narrative is too good not to go for least a Minisend imo. Also idc if we find a new third one or fourth lol. 3HqeWi1bCaffkMNnGpozoUh1o2VJtqmUzVJHeLBfpump GmGR4KUh4GQcKJJqom8Q7mbg5eSqCGu1sEx9krBpump
2
159
$toy coin appears to be making resurrection attempts after spending the night sub 4k as $food coin clears 1m and beyond. GksPeb4TjgkcLorASBFH1MxnvE5EpEb55usE46GZpump Keeping a close eye.
1
134
Shadow Zebra retweeted
We are excited to announce that next month marks the rollout of live demonstrations showcasing our latest technical milestones. To commercialize our deployment pipeline, we are shifting from isolating the conformal gate to offering a fully integrated inference solution. Clients can now deploy our embedded solutions directly on a pay-per-inference basis. Launching in collaboration with two initial partners, this offering is backed by our internal benchmarking and empirical validation studies to guarantee optimal model performance and tight uncertainty bounds. Stay tuned, the future of inference is imminent.
7
9
2
32
1,333
Been keepin an eye over here they been cookin. $RLAN CTyEzEC2WwUgNivmkSp6ZdqnPmBb59EyY4QmCXmFAJiy
This quoted post is unavailable.
1
1
83
I do love a good cook. I’ve seen this man cook months before, building for the joy of building. Plus he legit has apps on the iOS app store. That level of follow thru is rare. Curious to see where cybercab.city/ goes. $cab 4rQXtAwM4NPgTojbS31eJqgpUKbXS9ACbGEeFUZJpump
how do yall like this artwork style?
1
2
76
Shadow Zebra retweeted
1/ 🧵 Least privilege × conformal inference. An idea from 1975 says what an agent may touch. A theory built in 1998–2005 says what a prediction may claim. An AI agent that holds tools needs both, and the two fit closely. A long thread, with the math and live calls 👇 2/ Saltzer & Schroeder, "The Protection of Information in Computer Systems", 1975: "Every program and every user of the system should operate using the least set of privileges necessary to complete the job." Its stated purpose: limit the damage of an accident or error. 3/ The same paper states fail-safe defaults next to it: "Base access decisions on permission rather than exclusion." By default there is no access. Their argument: a mistake in a system that grants by permission fails by refusing. That is safe, and it gets noticed fast. 4/ An older source: Dennis & Van Horn, CACM 1966. A computation runs inside a "sphere of protection" given by a list of capabilities (a C-list). Delegation can only narrow: a capability you pass on must be implied by one you hold. Authority never grows by being handed down. 5/ Hardy, "The Confused Deputy", 1988. A compiler held a license to write its stats file. A user named the billing file as the debug output, and the compiler overwrote it. "The compiler serves two masters and carries some authority from each to perform its respective duties." 6/ Miller, "Robust Composition" (2006), the Principle of Least Authority: Permission is what a program touches directly. Authority is every effect it can cause, via others too. "we must provide the right amount of authority just-in-time, not excess authority just-in-case." 7/ Mainstream platforms now do this by default: • AWS IAM: "By default, all requests are implicitly denied" (the root user aside) • GitHub Actions: name one token permission and "all of those that are not specified are set to none" • NIST SP 800-53 has a control for it: AC-6 8/ Then came LLM agents 🤖 OWASP Top 10 for LLM Apps 2025, LLM06 Excessive Agency: "The root cause of Excessive Agency is typically one or more of: excessive functionality; excessive permissions; excessive autonomy." Mitigation 7: authorize downstream, not in the LLM. 9/ Simon Willison named the "lethal trifecta" in June 2025: private data, untrusted content, and a way to communicate externally. If an agent has all three, an attacker can make it send your data out. Least privilege gives the fix: remove one leg. 10/ Research builds it in: • CaMeL (Google, Google DeepMind, ETH Zurich) tags values with capabilities: 77% of AgentDojo tasks solved with provable security, 84% undefended (v2) • Progent checks a least-privilege policy on each tool call: attack success 39.9% → 1.0% (v3) 11/ Both state limits. CaMeL: "No, prompt injection attacks are not fully solved." Progent "cannot be used to defend against attacks that operate within the least privilege". Least privilege bounds what CAN be done. It says nothing about whether the action chosen is right. 12/ That second question is statistical: how often does the agent's claim miss? 📐 Conformal inference answers it without assuming the model is right. Vovk, Gammerman, Shafer and colleagues built it in 1998–2005; the fast split form is Papadopoulos et al., 2002. 13/ Split conformal in four steps: • fit any model on one dataset • score n held-out points, s_i = s(x_i, y_i), larger = worse fit • k = ⌈(n+1)(1−α)⌉, q̂ = the k-th smallest score • C(x) = {y : s(x, y) ≤ q̂} The set holds every answer the data do not rule out. 14/ If the n calibration points and the test point are exchangeable: P(Y ∈ C(X)) ≥ 1 − α and, with no ties, ≤ 1 − α + 1/(n+1). Why: the test score's rank among the n+1 scores is uniform. No correct model, no parametric family, no asymptotics. 15/ A small n limits what you can ask for. n = 10, α = 0.1: k = ⌈9.9⌉ = 10, so q̂ is the largest score. n = 10, α = 0.05: k = ⌈10.45⌉ = 11 > n. The textbook answer is q̂ = +∞, a set holding every label. The smallest α that n points allow is 1/(n+1). 16/ What it does NOT say: • Coverage is marginal, averaged over calibration and test draws. • Coverage given every x is impossible without assumptions on the distribution (Vovk 2012; Lei & Wasserman 2014). • It is never the probability that THIS answer is right. 17/ The marginal number also varies between calibration sets. Given one set, coverage is Beta-distributed near 1 − α. Angelopoulos & Bates: with n = 1000 and α = 0.1, it typically lands between .88 and .92. The guarantee belongs to the procedure, not to one call. 18/ For groups: Mondrian conformal (Vovk, Lindsay, Nouretdinov, Gammerman, 2003) calibrates each declared category on its own points, so coverage holds per group. The price: each group lives on its own n. Groups are fixed in advance; a thin group gets nothing useful. 19/ Two edge cases are signals: • The full set is "certain to be correct" and tells you nothing (Shafer & Vovk 2008). • The empty set rejects every candidate. It is a prediction known to be wrong, and a hint that the input looks unlike the calibration data. 20/ When exchangeability breaks: • Barber, Candès, Ramdas, Tibshirani (Ann. Stat. 2023): coverage ≥ 1 − α minus a weighted total-variation gap you cannot observe. • Gibbs & Candès (NeurIPS 2021) adapt α_t: a long-run error-rate bound, no claim per step. 21/ Beyond "is y in the set?": • Conformal risk control (Angelopoulos et al., ICLR 2024): E[loss] ≤ α for any bounded monotone loss. • Learn then Test (Angelopoulos et al., Ann. Appl. Stat. 2025): if no setting can be certified, it abstains. 22/ Now side by side 🔗 fail-safe default ↔ abstain unless the data justify acting least set of privileges ↔ smallest set the data allow grant per role ↔ coverage per declared group "grant everything" ↔ the full set revoke on change ↔ an exchangeability alarm 23/ The last row is real math: exchangeability martingales (Vovk, Nouretdinov, Gammerman, ICML 2003). Bet against exchangeability at fair stakes, starting from 1. By Ville's inequality, P(wealth ever reaches C) ≤ 1/C. A valid alarm that the ground under a grant has moved. 24/ Complements, not substitutes. Least privilege bounds the damage of an action, and says nothing on how often it is wrong. Conformal bounds how often a claim misses, and says nothing on the damage. An agent gate needs both: a ceiling on harm, a rate of error. 25/ Papers already sit where they meet: • KnowNo (Ren et al., CoRL 2023): conformal sets decide when a robot planner must ask a human. • A July 2026 arXiv paper gives LLM tool calls a risk budget per argument role: untrusted content "should never set a recipient." 26/ That is least privilege per argument, with a guarantee on how often each check fails. At MONARK, this is the design space we build in. Here is how the two halves meet in one gate 👇 27/ MONARK's gate decides, never acts. Docs: "The gate emits a decision; it never calls the tool it names. It holds no key, moves no fund". An agent sends a claim, gets commit, defer or abstain, an auditable region and the budget left. Never a probability of being right. 28/ Least privilege applies to the decider itself 🔒 Exactly four tools are served: attest, gate, cascade, calibrate. Their code imports no file-system, network or process module, calls no fetch and reads no clock. A test scans every tool file and fails on any hit. 29/ The caller cannot widen its own guarantee. On a calibrated class, the server sets α and nMin. Live calls today: α = 0.1 instead of 0.01 → 400 policy_alpha_mismatch nMin = 10 instead of 100 → 400 policy_nmin_mismatch Least privilege, applied to statistical parameters. 30/ Nor can it impersonate a calibrated population. Your own calibration sent on a calibrated key → 400 byo_overrides_committed. Look-alike keys (case, l/I/1, rn/m…) are refused too. The server derives the stratum from ŷ; the caller never picks it. 31/ Authority per category, default deny. The liquidation class is split into strata of ŷ fixed in advance; only stratum 0 is calibrated (n = 170). Live: ŷ in stratum 0 → commit, upper bound [0, ŷ + q̂] ŷ in stratum 1 → abstain, under_calib, no region 32/ Honest about small n: at α = 0.01 and n = 170, k = ⌈171 × 0.99⌉ = 170 = n, so q̂ is the largest score seen, reported as is. One more nine (α = 0.001) would need n ≥ 999. The served text says it: one recorded episode, no coverage claimed for any other event. 33/ Bring your own scores and watch the deny branch 🧪 Ten scores from 0.1 to 1.0, α = 0.1, so q̂ = 1. Live: A 0.85, B 1.5 → C = {A} → commit A 1.5, B 2.0 → C = ∅ → abstain A 1.5, B 0.5, intent A → C = {B} → abstain A 0.3, B 0.6, τ = 1 → C = {A, B} → defer 34/ Same scores at α = 0.05: k = 11 > n = 10. Where the textbook returns the whole space, the gate refuses: abstain, under_calib. Budget 0.01 under a floor of 0.05 → abstain, budget_exhausted. NaN, a zero-width region, an empty set: abstain. Nothing defaults to commit. 35/ The caller carries the budget; the gate does not hold it. The gate compares it with the caller's floor and hands it back unchanged. Each request gets a fresh server: no account, no stored caller data. A decider with no state has nothing to escalate. 36/ Keys 🔑 In the data path, exactly one module may read a paid endpoint key, and the key may only travel to a closed list of exact hosts, over https. Every other module names providers by label. A run opens and locks only the providers it names. 37/ Separation of privilege ("two keys") on the evidence side: our on-chain sensor reads value-bearing data from two distinct providers. If the bytes differ, the window fails closed. In CI, the workflow token is read-only; every action is pinned to a full commit SHA. 38/ Read `allow: true` correctly: it is a coverage verdict on YOUR prediction, not permission to run the tool. Your runtime enforces the act; for spend caps, use your platform's controls. The door every transfer, swap or signature would pass first is named, not delivered. 39/ Reading rules 📚 • Coverage is marginal, under exchangeability, and void under drift unless re-checked. • Abstain is an answer. An empty set is a signal. • The caller cannot widen its own terms. • Never a probability of being right. monarkgate.tech/docs/gate 40/ Sources: Saltzer & Schroeder 1975 · Dennis & Van Horn 1966 · Hardy 1988 · Miller 2006 · OWASP LLM06:2025 · Willison 2025 · arXiv 2503.18813, 2504.11703, 2307.01928, 2607.24343 · Vovk et al. 2003, 2005 · Barber et al. 2023 · Angelopoulos & Bates github.com/KraidleAI/Monark
3
6
24
779
Shadow Zebra retweeted
A signature that keeps repeating For years, we have watched the same pattern return, again and again, with almost perfect regularity. Before every major rise, whether you call it a bull run or a climax run, wealth concentrates first. We first saw it on Bitcoin, nine years ago. Since then, we have seen it repeat on every crypto asset that went on to make 100x or more. How major trends are born Deep trends do not start by chance. They are set off by raids: large wallets moving with intent. They are then fed by hundreds of thousands of smaller wallets, once the evidence becomes impossible to ignore. What this demands of an agent To act on this, an agent must locate itself every day, on every asset, both fundamentally and technically. Only then can it calibrate its strategies across the short, medium and long term. A take-profit is not set on price alone. It is also set on timing, on a window of time. This means seeing the whole picture: knowing when, how, why, where, and who. Where we stand Everything you see of Monark today will be nothing more than the docs page of the Monark we are building. We are not tweeting much these days. We are building a great deal. Monark has fewer than thirty commits: github.com/KraidleAI/Monark Monark Governance, our workshop, has more than 3,500: github.com/KraidleAI/monark-… The work never stops. We are building the future. We will also open-source the Shogen repository, once it has been cleaned up through a final set of gates.
1
4
28
649
Iiiii feeeel like solana:4c1XZRqFV6y8pAckHru5oiGPYUFw1eQ3kFotPLHrpump should be a bit higher than 20k no? 4c1XZRqFV6y8pAckHru5oiGPYUFw1eQ3kFotPLHrpump
We went quiet for a week. Here's everything that happened. KeepHost is a shielded pool on Solana. You put in a fixed amount, you take it out later at a different address, and nothing on-chain connects the two. Not because we hide the link because there is no link to hide. What's live, right now, and checkable in one click: The program is on mainnet: CTHg29kf7L6TNDH5TSd3tdoZfsmP39JjyQWKmPtEY1YW Three pools are open 0.1, 1 and 10 SOL. Fixed sizes, on purpose: an amount that's unique to you identifies you no matter what else you do. The vaults are PDAs. There is no private key to them, and no admin withdrawal in the code. That's not a promise about our intentions — it's an absence you can verify. The relayer pays the fee on your withdrawal so you don't need a funded wallet to receive. It takes 0.002 SOL. The proof makes it mathematically unable to redirect your payment or raise its own cut, and the CI proves that on every commit. What we shipped this week: The round trip now runs entirely in a browser deposit, receipt, proof, withdrawal. No terminal, no install. The site is a dashboard now instead of a pitch. Every number on it is read off the chain. A wallet scanner. Paste any address and see what it tells a stranger about you. It's free and it's not really our product — it's the reason our product exists. The numbers, honestly: 7 deposits and 2 withdrawals so far. 5 sitting in the pool right now. They're ours. We're not going to dress up our own test transactions as traction. An anonymity set of 5 is an anonymity set of 5, and it's the one number here that is simultaneously the marketing and the product: every deposit makes every other deposit harder to trace. Small pool, weak privacy. That's just true, and pretending otherwise would be the one lie that actually hurts someone. What isn't done, before anyone has to ask: No external audit. None. The proving key still comes from a setup run on a single machine. The ceremony that replaces it is open two minutes, nothing to install, and one honest contributor is enough to close it permanently. The upgrade authority is still held, deliberately, until that key ships. It sits on an offline wallet, so taking the server no longer means taking the program. Association sets proving your deposit belongs to a public, reproducible set without revealing which one are written and tested. Not deployed. We'll say so the day they are, and not before. Until that key ships: deposit what you can afford to lose. Most projects tell you that afterwards. Everything above is in SECURITY.md and CEREMONY.md. The source is public. keephost.fun
1
1
2
245
Must. Accumulate. solana:FYZcYCHSp8FzNba1UtDZydKKGosmxVNpFBiVuia38AhT *mantra for my last week+ FYZcYCHSp8FzNba1UtDZydKKGosmxVNpFBiVuia38AhT
2
181
👀👀 $anima been cooking I see. Been keeping watch. CQK42yTEyKe2QdEcBW7NK3Wa1tv4goEPHUFZhAwSAte6
Project spotlight #3 is Anima. Anima (@AnimaAgent) is a Solana research-and-swap agent with a 3D face and voice: it streams answers, runs live market tools, and returns unsigned swaps sealed in expiring, wallet-bound quote tokens. Fund the agent, share it, no seed phrase.
1
231
Shadow Zebra retweeted
📍 Day 1 is on the board. 🔗 Release: github.com/KraidleAI/monark/… 📡 Live tracker: monarkgate.tech/narabi Narabi took its first real step last night at 00:39 UTC. One day of attested stablecoin flows, evaluated, hashed, and published: 🗓️ Day 2026-09-18 · evaluable 🌊 Regime: calm 📏 Bound: 2 🔐 Digest: 9b5f89fd No dashboard opinion, no score. A daily line in a public timeline, each one chained to the previous, that anyone can recompute from the raw attestations. What it does not do: it never tells you it is right. It tells you what it saw, what it refused to evaluate, and how far it is from the coverage target (1789 days on today's projection, so yes, this is a long game ⏳). Day 2 lands at 00:39 UTC tomorrow. Same file, one more line. 🧵
5
13
26
1,892
Always nice to see day 2 exists. $vortex 👀👀 7CXJcHs14RSn5ftub65ET8D6yRexppeacXnd9w44pump
Late GM guys. Working on some updates right now. I'll update you as soon as they're live.
1
2
218