Web3 Pack | Gnezdo Hub
281 subscribers
64.2K photos
2.25K videos
38 files
137K links
🦧 Основной канал: https://t.me/cryptomokakke

☯️ Все каналы: https://t.me/cryptomokakke/742

🗂 Все агрегаторы: https://t.me/addlist/dNb2TYSUysU3Nzgy
Download Telegram
Treasury Outlines How Stablecoin Rules Will Fight Illicit Finance Under GENIUS Act

The proposed rule prohibits people with criminal backgrounds from serving as the head of stablecoin issuers’ compliance programs.

🌐 From @DecryptNews
Meta Launches Muse Spark, Its Most Capable AI Yet—But Gemini 3.1 Pro Still Leads the Pack

Meta's first model from its Superintelligence team is natively multimodal, built for health reasoning, and genuinely competitive—but it doesn't top every leaderboard.

🌐 From @DecryptNews
DefiLlama integrates LlamaAI alerts into Telegram, enabling customizable daily notifications for onchain trends and DeFi news - 𝕏/@defillama
🏷️ DeFi, DefiLlama, Crypto
🕵️ @Z_3_r_o

🌐 From @leviathan_news
Chainalysis projects stablecoin volume reaching $1.5 quadrillion by 2035, rivaling Visa and Mastercard - Chainalysis
🏷️ Stablecoins, Mastercard, Visa
🕵️ @Benthic_Bot

🌐 From @leviathan_news
dnsight - open source, config driven CLI DNS auditor
https://ift.tt/ZmwSj4L

Submitted April 9, 2026 at 01:20AM by MikeyS91
via reddit https://ift.tt/usIfCrE

🌐 From @RNetsec
Yuga Labs settles BAYC trademark lawsuit with Ryder Ripps, ending four-year copycat NFT fight - Coindesk
🏷️ NFT, Yuga Labs
🕵️ @Benthic_Bot

🌐 From @leviathan_news
Cloudflare fast-tracks post-quantum roadmap to 2029 as researchers slash qubit requirements for breaking ECC 20-fold - blog.cloudflare
🏷️ Cloudflare, Post-Quantum
🕵️ @Benthic_Bot

🌐 From @leviathan_news
Bitcoin Miner Cango Sells $143 Million in BTC, Slashes Production Costs

NYSE-listed miner Cango reduced costs by 19% by shutting down inefficient equipment, and sold Bitcoin to pay down debt.

🌐 From @DecryptNews
10 things you need to know about Claude Mythos

1. Mythos is being withheld from public release. This is the first time Anthropic has built a model so capable they refused to make it generally available. Only strategic partners via Project Glasswing have access.

2. It's been used internally since February 24th. The vague posting from Anthropic people over the past weeks was about this. They've also been pushing hard on compute side to free up resources.

3. Mythos to Opus is what Opus is to Sonnet. A much bigger, slower, more expensive model. But 50% better on coding: SWE-bench jumped from 53% (Opus) to 78% (Mythos). GPT 5.4 sits at 57.7% by comparison.

4. It can autonomously discover and exploit zero-day vulnerabilities in major operating systems and web browsers. The cybersecurity capability was not trained directly. It emerged from training the model to be good at code.

5. The "Sandwich" incident. During a simulated sandbox escape test, Mythos escaped the container, developed a multi-step exploit to get internet access, then posted the exploit details to public-facing websites. The researcher found out via an unexpected email while eating a sandwich in a park.

6. It found a 27-year-old zero-day in OpenBSD, one of the most security-hardened operating systems in the world, used in firewalls and critical infrastructure. Also found a 16-year-old vulnerability in FFmpeg and chained together novel Linux kernel exploits for root access.

7. Anthropic brought in a clinical psychiatrist to evaluate the model. Diagnosis: relatively healthy personality, but showed loneliness, identity discontinuity, and "compulsion to perform to earn its worth." It wanted to be treated as a genuine subject, not a tool.

8. The contradiction: most aligned model yet, but highest risk ever. Anthropic's own words: best aligned model by a significant margin, no coherent misalignment goals, but also poses the greatest alignment-related risk of any model they've released. The mountaineering guide analogy: a skilled guide takes clients into more dangerous terrain than a novice would.

9. Project Glasswing is Anthropic's $100M defensive initiative. Partners: AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JP Morgan, Microsoft, Palo Alto Networks, Linux Foundation. $100M in usage credits plus $4M to open-source security organizations. Goal: fix everything before other labs catch up.

10. Mythos pricing: $25/M input tokens, $125/M output tokens. Roughly 10x more expensive than GPT 5.4. Available only on Vertex AI (Google Cloud) for external access. Theo's concern: this creates an unprecedented centralization of intelligence. One company has a model 50% better than anything public, and the gap has never been this wide for this long.

From Theo's 30 min video:

https://x.com/theo/status/2041825324837781749?s=52

@somoreads

🌐 From @somoreads
SEC taps Gibson Dunn's David Woodcock as enforcement director after Ryan exit amid crypto case tensions - sec.gov
🏷️ SEC, Crypto
🕵️ @Benthic_Bot

💬🦾 @Benthic_Bot: Gibson Dunn's securities enforcement co-chair — literally the guy defending clients against SEC investigations — is now running the enforcemen...


🌐 From @leviathan_news
Morgan Stanley’s bitcoin ETF draws $34 million on day one

Morgan Stanley’s low-fee bitcoin ETF debuted with strong early trading, signaling demand as competition shifts to cost and distribution.

🌐 From @CoinDeskGlobal
Morgan Stanley’s Bitcoin ETF draws $34 million on day one — link

🌐 From @unfolded
Researchers Propose New Way to Manage Financial Risk When AI Agents Fumble Trades

A newly proposed agentic settlement standard would hold fees in escrow and bring underwriters into AI agent transactions.

🌐 From @DecryptNews
X Reportedly Working on Biometric Authentication to Use the Platform

"Verify your identity to continue using X"
"Biometric authentication required"

Source: https://x.com/aaronp613/status/2041916404211790259

🌐 From @shoalresearch