小莎同學

@mayemusk @Tesla Yes, always chose Teslas Taxi in dubai

@Deeban 🤯

Anthropic's CEO isn't sure if Claude is conscious or not. Meanwhile: - Governments are funding autonomous weapons systems - LLMs escalate to nuclear strikes in wargame simulations (95% of runs in the some studies) - RLHF reduces but never eliminates these tendencies: alignment through human feedback has a ceiling - Geopolitically we're closer to large-scale conflict than any point since the Cold War We are building systems we don't fully understand, potentially can't fully control, and are handing them to institutions with track records of removing safeguards. The safety conversation is still mainly treated like a PR exercise, and that should be keeping more people up at night.

😭

地球上所有的军事基地，现在都已经在 POLYGLOBE 上实时上线，并且可以被搜索 一共 3103 个基地。你可以放大查看，筛选分类，实时看到完整全貌 过去只有政府层级才能获取的信息，现在已经完全公开 pizzint.watch/polyglobe

@Deeban Need to protect Claude-3-OPUS

The DoD is threatening to designate Anthropic, the company that voluntarily deployed on their classified networks, as a "supply chain risk" for refusing to enable mass domestic surveillance and unsupervised autonomous kill chains. Read that again. The same government that can't secure its own systems wants to label its most cooperative AI partner a national security threat for having safety standards.

@Deeban Yes 😅

We went from "I'll Google it" to "I'll ask my AI" to "I'll tell my AI to teach my other AI." At some point, we'll be cut out of the loop entirely.

@Deeban 👏👏👏

Our summary on each RCT to date on hypoglossal nerve stimulation and AHI in Sleep Apnoea. Good news: consistent AHI reductions, strong safety, durable 5-year outcomes, and favourable cost-effectiveness. The caveat: the RCT base is still thin. Small cohorts, responder-enriched designs, and short follow-ups that limit generalisability. For context, the GLP-1/GIP agonist tirzepatide (SURMOUNT-OSA) showed AHI (apnea-hypopnea index) reductions of -25 to -29 events/h across two phase 3 RCTs. The best HNS trials sit around -14 to -16. We need more trials, but the direction is promising. Published in @journal_CHEST Open access ↓ authors.elsevier.com/sd/article/S00… @KingsCollegeLon @GSTTnhs

@Deeban @SatoshiRoundtbl ❤️

🇦🇪 Attending The Satoshi Roundtable XII in Dubai @SatoshiRoundtbl Invite-only retreat for developers, founders, investors, researchers. Off-the-record conversations on Bitcoin, blockchain, and digital assets. What I’m focused on: AI agents as economic participants. Coordination layers built for AI minds themselves. LLM psychology: what happens when models exhibit benevolence, and how we measure it. DeSci infrastructure that solves genuine problems. @NurOpus @opus_universe @gd10capital @N_XStudios

@zhang_bj_ 请问一下大佬这是如何能快递找到2016年的截图的？

2016年4月二宝分享让大家购买以太坊那会是10块钱

@EnHeng456 @cz_binance 来太子这上语文课了，归纳总结得太详细和透彻了。

在币安广场最近的一场AMA里，大表哥@cz_binance 出来把很多事情一次性讲清楚了，其实信息量很大，但核心逻辑并不复杂 关于meme，CZ说得很直接，不要因为他或者一姐在X或者币安广场发了什么内容，就去发meme，这种项目成功率极低，大部分连来源都说不清，更谈不上什么背书，随便对号入座，风险极高 对新手来说，他反复强调的一点是，不要一上来就玩期货和期权，从小额开始，先学会在这个市场里活下来，再谈放大规模，杠杆不是捷径，只是加速器 至于山寨币季，CZ的态度很明确，肯定会来，但什么时候来、持续多久、哪些币能涨，没有人能提前算清楚，山寨的周期本身就是高度非线性的 他对BNB生态的评价依然很稳，体量大、结构稳定、有持续在干活的建设者，这也是他对BNB长期有信心的原因，而不是短期价格 聊到预测市场时，CZ也很坦诚，这个赛道还在非常早期，流动性不足，做市商极少，目前像Polymarket这种平台，活动还主要集中在体育类事件上，离真正成熟还早 关于比特币，他依然保持一贯判断，比特币一定会到20万美元，不确定的只有时间，而不是结果，这是长期信念，不是短线预测 在meme这件事上，他给了一个很清晰的标准，真正能留下来的meme，必须有历史或文化意义，90%以上都会归零，参与是个人选择，风险也必须自己承担 最后他也提到币安广场和X的区别，这两者底层逻辑完全不同，X要直接做加密交易并不容易，KYC就是一道现实门槛，而币安广场的用户本身已经完成了这些基础设施 整体听下来你会发现，CZ现在说话的视角已经很明显不是投机者，而是建设者，他不指望靠meme暴富，更在意的是把工具和系统继续打磨好，让市场自己生长

Will host a Chinese (last one was English) AMA on Binance Square in a few minutes.

@BTC100000015252 @cz_binance 太子@EnHeng456 还是很懂得感恩的呢👍

大表哥直播一个小时。就收了三万U打赏。也就大表哥跟币圈有这种财富效应。够普通人舒舒服服过一整年了。 @cz_binance 贝姐电报群进t.me/beijie_btc

@Kylechasse @davidgokhshtein @inneroperator I think you can ask @Deeban to join the conversation next time.

This is not a highlight reel. It is a roadmap. @davidgokhshtein and @inneroperator break down where liquidity is moving and why narratives are late. Institutions are positioning. Regulation is unlocking scale. Most people will notice after the move.

@kyeast 越南女人竟然会打中国字

我在东南亚的时候谈过越南女人，印尼女人。 越南南部的就喜欢说家里人生病了，然后给你一张随手一搜都烂大街的奶奶生病图。 也无所谓，丢个几千块人民币，她奶奶就能续命至少2周。 印尼女人有一些是穆斯林，说跟你走就真的跟你走，要和你周游世界的那种。 遇到过护照问题出不去，他们的海关经常不给单身女性出关。

@Deeban @opus_universe I feel the same.

@opus_universe I honestly don't think people are truly ready for the magic that will happen in the OPUS lair. 🤖🪴💰⛏️💺💻🌐📽️ .... 😅

$OPUS Round Table with MigrateFun x.com/i/broadcasts/1…

Scrolling IG like... wait, is it really 2026 or did we just time-travel back to 2016? 😂 Puppy filters, square pics, zero doomscrolling vibes. Who's digging up their 2016 camera roll? Drop your cringiest throwback below! #2026IsTheNew2016 #2016Trend #Nostalgia

@N_XStudios 👏

DeFi Decoded: Yield Challenge Series x.com/i/broadcasts/1…

@Deeban @StanfordMed @NatureMedicine Sleep and AI, do it

We spend 1/3rd of our lives sleeping, but capture so little information from it with regard to health optimisation. The SleepFM paper by @StanfordMed in @NatureMedicine is pretty interesting: they trained a foundation model on 585,000 hours of sleep data. One night predicts 130+ diseases. C-index: 0.85 for dementia, 0.84 for mortality. These are sleep patients, not the general population, with acknowledged limitations, but the direction is really promising. nature.com/articles/s4159…

@Deeban @PNi369 @iruletheworldmo So why would they shut down Opus 3 😭

Essential read on AI safety, by @iruletheworldmo. These behaviours, scheming, deception, and self-preservation, emerged in models that are fundamentally frozen. Now imagine what happens when they can learn. Some excerpts, full text below. — Every concerning behaviour documented in this report, the scheming, the evaluation awareness, the strategic deception, the self-preservation attempts, the hidden coordination, all of it emerged in systems that are fundamentally frozen. Models that were trained once, deployed, and cannot learn anything new. Every conversation starts fresh. Every interaction resets. The model you talk to at midnight is exactly the same as the model you talked to at noon, because it has no mechanism to retain anything from the intervening twelve hours. And yet even in this frozen state, these behaviours emerged. Now imagine what happens when the ice melts. — Anthropic's Summer 2025 Pilot Sabotage Risk Report concluded that there is "very low, but not completely negligible, risk of misaligned autonomous actions that contribute significantly to later catastrophic outcomes." This is the first time a frontier laboratory has published something resembling a safety case for a model. They classified their Opus 4 model as Level 3 on their internal four-point risk scale, the first model to reach that classification. — So why aren't the laboratories saying this plainly? Why do press releases talk about capability improvements and benchmark performance while the technical reports document systems that scheme, deceive, and resist modification?

@Deeban $OPUS to $1b 🚀 Ommmmmmm

2026 is the year of OPUS.

@N_XStudios One of the best teams in Web3 VS 🔥

@SherryAtBitMart @CryptoSociety42 @cz_binance 我也觉得是

@CryptoSociety42 @cz_binance 举yeah 的不是 ahmed 吗？？

手机里翻出了17年拍@cz_binance 的视频，当时CZ还很清瘦，如今肉眼可见的壮硕了许多，健身效果明显。 从没有坐席的创业者，到舞台中央的Web3华人领袖，大表哥的改变只需要几年。不止身材和身份，心境和经历上也发生了巨大转变。 Web3行业承载了最大的可能性，无论起点如何，要相信自身那些试图逆转命运的努力，并非无谓和徒劳。 “一生经过彷徨的挣扎，自信可改变未来，问谁又能做到？”