고정된 트윗
Sean
2.5K posts
Sean
@WtsSeanBuilding
Co-Founder of https://t.co/PIO7eil8Nc & @trypeony | @Imperialcollege @UKDRI among other ML and Eng roles
Austin, TX 가입일 Ağustos 2020
960 팔로잉1.2K 팔로워
I’ve been to UBTech’s showroom many times. It’s pretty cool. Once our Shenzhen Robot Valley showroom opens next month, feel free to drop by too.
UBTECH Robotics@UBTECHRobotics
Great conversations at UBTECH today with students from @uclaanderson Global Immersion Program!🤖🌏 From next‑gen humanoid robots to the real‑world business of embodied AI — inspiring to share where robotics is headed with future global leaders. Thanks for visiting, and welcome to Shenzhen.🤝 #UBTECH #UCLA #WalkerS2 #Humanoidrobots #Robotics
English
@yieldsearcher Wait explain to my smol brain are you talking about how this currency move is a preventation to what might happen to China that have happened to Korea before?
English
Chinese leaders take a long view, and know what happened in Korea.
Its military dictatorship collapsed in 1987 at its peak middle class growth.
Parents gen stayed loyal to the junta for delivering prosperity; their college-aged kids, empowered by it, fought against them.
(Obv there is more nuance to it than what I can fit within 280 limit, but this generation divide still runs the current Korean political current to this day, even when these voter bases are now at their 80s and 60s respectively.)
Mr. VIX@yieldsearcher
Inevitable, but first China must liberalize the currency and capital controls, allow FX appreciation, and be willing to run current-account deficits for the RoW. Most importantly for CCP, it must brace for a booming wealthy middle class less reliant/obedient to the state.
English
[ Perp DEX 2026 ]
2026 的第一天，整理了一下各个 Perp DEX 的发币预期，用来给 2025 年留下来的一堆积分做一个相对理性的估值参考。
FDV 全部根据 Polymarket 的数据计算（用钱投票最准确）：
把「FDV 高于某个数」的概率视为累计分布 →
用相邻档位做差得到各 FDV 区间概率 →
再用区间中点做概率加权平均，得到市场隐含的「预计 FDV」。
表格按 剩余可获得积分的周数 排序，方便横向比较。
数据不代表任何个人意见，仅仅是根据所有公开信息，为我自己手上的积分价值有个大概判断，顺便整理出来分享给大家。
如果有任何数字或信息不准确，欢迎指正 🙏
* grvt 没有上 polymarket，所以估值取了 paradex 和 extended 的中间值。
中文
@yourQuantGuy 实际上预计空投比例是不准的, 就像Lighter说的25%实际上就10%.
根据理性经济学人假设来看, 目前的OTC价格就是市场公允价格. 比如说 Paradex 每分 OTC现在大概 0.2U. grvt 大概 7U.
中文
@chenreason 好像市场这么反应的一部分原因也是因为net import drop 被算为gdp增长，并且consumer spending揉开看最大的增长来自于healthcare insurance，代表实际经济一般
中文
20251224 美股市场观察，🧑🎄
1，昨晚标普 500 再创历史新高，纳指表现也很强劲，成长股跑赢价值股，但是代表中小盘的罗素 2000 指数下跌 0.7%。所以虽然大盘在涨，但手中有中小票的朋友其实昨晚在滴血，圣诞行情里赚了指数却不赚钱。
2，指数虽然新高，但昨晚标普 500 成份下跌家数略多于上涨家数，广度一般，只有35 家 52 周新高，且有5家新低。走强的方向主要是超大市值科技、通信服务和部分医疗保健，一些防御板块和传统周期板块略走弱，小盘股整体净流出。
3，昨晚 GDP 数据爆表，美国三季度 GDP 修正值高达4.3%，预期只有 3.3%，创2023年第四季度以来最高增速。这数据太强了，强到让多头和空头都沉默了。好的一面是经济衰退的鬼故事彻底没人讲了；坏的一面是，如此火热的经济让美联储失去了在一月降息的理由。
4，除了GDP 数据，昨晚的个人消费支出数据 (3.5% vs 预期 2.7%)， 也大幅超出预期，消费非常火热，但12月的消费者信心指数从92.9降到了89.1，连续第五个月下滑，是关税开征以来的最低值之一。总结起来就是，数据告诉我们消费很强劲，但情绪其实让人有些焦虑。
5，昨晚数据出来后，10年期美债小破了一下4.2%，现在又回落到了4.16%附近，在经济高增速的背景下，长债只是小幅抬头且回落，说明债市并没有定价二次通胀失控，更多是接受一个更高的中性利率。
6，目前市场对联储1月议息会议被动维持利率不变的概率预期回到了87%左右，比前两天的73%提高了不少，市场现在把首次降息预期推到了2026年6月附近，但依然计价2026年会有两次 25bp 降息。市场从很快就降息的预期调整为利率高位停更久，但最终还是会降的预期，这对风险资产而言，不是最理想，但也不是灾难。
7，在这么靓的 GDP 之下，尽管美债收益率维持高位，美元指数却诡异地崩到了 98 以下，创 10 月以来新低，并且今年累计跌幅接近 -9.6%，或将是 2017 年以来最差一年，昨晚的黄金、白银、铜全线创历史新高。
8，纽约联储在最新文章里确认，FOMC 在 12 月会议上取消了“常备回购工具（SRF）每日 5000 亿美元总额度上限”。简单介绍下SRF，就是当金融机构手里有大量国债等优质债券，但暂时缺现金，或市场短期利率有上涨压力时，机构可把债券卖给美联储，同时约定未来按固定价格买回来这就是回购。本质是美联储给市场临时输血，帮机构缓解短期资金紧张，进而稳住市场利率。
最简单的理解就是，联储从原本的有限送水救火，现在变成它会无限兜底市场流动性，给市场一颗定心丸。
9，再看看恐慌指数， VIX 已经跌至 14 附近的年度低位，市场目前处于一种无所畏惧的极度贪婪状态，CNN 的情绪指数同样来到贪婪区间。
10，差点忘了今晚 21:30 还有一份初请失业金人数的数据要发布，预期22.4万，这个大概率是今天只有半天交易时间的主要波动来源了。
11，最后给两组最近发布的数据，
- 周一那晚股票类 ETF 再平衡净流出高达 398 亿美元。
- 2025 年散户交易占美国股票交易量的 20–25%，高的时候能到 35%，创历史新高；散户偏好 AI、特斯拉之类的故事股，和主题 / 杠杆 / 加密相关 ETF。
--
平安夜快乐朋友们！🧑🎄
对冲积鲸Reason@chenreason
20251223 美股市场观察， 1，昨晚圣诞行情开始，美股呈现出假日漂移模式，三大指数集体收涨，标普500逼近6900点历史高位。只不过成交量因圣诞假期临近而显著萎缩，低于近 20个交易日均值14%。市场在节前维持敢冒险，但不想搞大事的姿态。 2，虽然成交量偏低，但市场结构不算太坏。多数板块上涨，risk on 启动。罗素2000跑赢了纳指和标普，NYSE的涨跌家数比约为2.1:1。昨晚资金没有完全抱团在七巨头取暖，而是开始向中小盘和周期股扩散，资金明显从公用事业等防御板块流出，涌向科技、半导体和银行，金融板块创历史新高。 3，恐慌指数躺平，VIX创年内新低，跌破15，收在14.08。期权市场显示看跌保护的需求骤降。这虽然是多头的狂欢，但也是潜在的隐患，低波动率会诱导风险平价基金上杠杆买股票，这在短期是助燃剂，但一旦有风吹草动，拥挤交易的踩踏会非常惨烈。当所有人都不觉得有风险时，风险往往就在拐角处。不要在这个位置裸卖期权去赚那点权利金，盈亏比极差。 4，别只看股市涨，看看债市。10年期美债收益率悄悄摸到了4.17%，整体还是略高不低，在降息预期下收益率不降反升。反观30 年美债收益率，目前4.8%左右，全年几乎没涨也没跌，但放在过去一年的整体环境里来看，还是很离谱的。 毕竟过去一年， - 2025年金价暴涨70%， - 科技股经历25年一轮牛市， - 美联储全年降息75bp， - 美元指数跌约10%，通胀持续高于目标， - 财政赤字因为新一轮减税计划扩大， - Fed 独立性受到政治压力。 5，在股市 risk on 的同时，黄金和白银双双刷新历史纪录。通常避险资产和风险资产是跷跷板，现在却是齐飞。美联储偏鸽，市场很鸽，贵金属极度鸽。按最新测算，白银总市值约3.7 万亿美元，全球排第四，仅次于黄金、英伟达和苹果，已经超过Alphabet。上方空间更多由制度不确定+地缘升级驱动，很难预测；但下方一旦出现美国数据意外强、地缘缓和、或央行释放金价过热等信号，回撤幅度也不会小。 6，今晚 21:30 有美国第三季度实际GDP终值、耐用品订单、消费者信心指数公布，GDP终值预估3.3%，前值3.8%。这基本就是今年最后一批对软着陆 vs 增长放缓有实质信息量的数据了。如果今晚结果显著低于预期，可能打击风险资产但继续强化降息交易；如果显著高于预期，那降息节奏预期可能会受压制，对长债不利，会加大贵金属 + 高成长股这类资产的拥挤度与泡沫争议。 7，个股方面，特斯拉盘中一度 +3.7%，收盘仍有约 +1.6%，股价逼近 500 美元整数关，主要催化来源于特拉华最高法院恢复马斯克 2018 年天价薪酬方案，消除了悬在特斯拉头上一大巨大利空（ x.com/chenreason/sta… ），且市场再度炒作 Robotaxi + Optimus 机器人的远期故事，情绪面再剧烈催化。 8，今天最后聊一嘴 crypto，这周五有超过236亿美元的比特币期权将要到期，是最近几年规模最大的到期事件之一，这次大仓位整体偏向多头，put 和 call 的比例大概是 0.38，押注价格上涨的人的多余押注价格下跌的，目前的最大痛点接近 96k 美元，请注意近期的可能大波动。
中文
FRED and Indeed both showing wage growth at its slowest pace since 2021.
Unlike the 1970s,
- Wage growth is much weaker
- Economy is much more wage/less energy-intensive (service is 80% from 65%)
- Older demographics with perhaps negative household creation
Hard setup for stag.
English
MiniMax 今天递交了港股上市招股书，
因为上周刚看完老罗对他们 CEO 的访谈，所以抱着一丝好奇心翻了下他们的招股书。
公司历史，产品，营收，各家媒体也都讲的比较透了。
唯一一个比较吸引我的点藏在了他们的风险因素里。
大意就是，如果因为各种原因没法获得英伟达的先进芯片，可能会对公司产生重大负面影响。
具体，见图，招股书第 92 页。
对冲积鲸Reason@chenreason
周末深夜，2倍速刷完了最新一期罗永浩的十字路口，访谈的是MiniMax CEO 闫俊杰。 说实话，这期信息密度不是很高，但在快结束时老罗问了个比较有意思的终极问题。 实现 AGI 后，还需要人机协同吗？人类浑身都是缺点和臭毛病，要人干什么呢？ 我暂停了视频，想了好一会，一时间真找不到人在 AGI 后存在的意义。 最后赶紧把问题抛给 GPT，没想到GPT 的回答让人心酸又好笑。 GPT回答道，人类的弱点就是太脆弱，脆弱的人类刚好是 AGI 实现后所有事物责任闭环的重要环节，因为 AI 不会感到痛苦，但人类会，这种痛苦的感觉让人类恰好成为最好的 AI 监督者。 因为没有人愿意背锅。
中文
@tig88411109 @Golf_Trump_ This won’t work, institutions have access to deeper level of info than retails if they want to, retail’s only edge is “intuition” on their understandings of people behind certain companies
English
Retail investor isn’t “dumb money” anymore. Retail is the tape.
Q3 2025, individuals were 20% of US stock volume, second only to the 2021 meme peak (hedge funds and long-only mutual funds, both around 15%)
It’s a market structure shift. But retail is taking over? Wrong framing.
Retail didn’t suddenly get smarter. Retail got louder. The edge didn’t disappear. It moved.
For 20 years the best retail strategy was PASSIVE indexing for one simple reason: you could not compete on research scale, access, or execution. So you rented returns via ETFs and let institutions do price discovery.
AI just broke that excuse.
Not by making you faster. By making you deeper.
Here’s the counterintuitive truth: retail will never beat institutions at execution.
If you’re playing charts, gamma, multi-leg hedges, and intraday “risk management,” you’re doing CrossFit against Olympians. You’re not “sophisticated.” You’re donating.
Retail’s only durable edge is the opposite of what retail chases:
attention and time.
Institutions are diversified by mandate. They cover too many names. Their attention is diluted by design.
Your advantage is you can pick five companies and actually learn the machine under the ticker: incentives, cost curves, competitive moats, cash flow reality.
Then the real leverage: time.
Funds have LPs, quarterlies, drawdown limits, career risk. They can be right and still be forced out. You have one risk committee: your sleep.
So the AI-era retail playbook is brutally simple:
Use AI to compress research friction.
Use human judgment to pick mechanisms, not narratives.
Use time to hold through noise when the thesis is intact.
Stop trying to be a mini hedge fund.
In this regime, retail doesn’t win by trading more. Retail wins by thinking deeper… and holding longer.
The market pays for mechanisms.
Noise is free.
Signal. Time. Conviction. Sleep.
English
was promised warm weather
so far:
> insanely cold
> insanely depressive rain
sf is cute but the weather has been so bad :(
English
@WtsSeanBuilding Deflation without a recession would be great for both asset classes.
English
Last Dec, with Core CPI at 3.3% and UR at 4.1% with everyone starting to get jittery on incoming Trump tariffs,
(1) Would you have guessed that we would be at 2.6% core CPI (3.0% if you go by Sep) and 4.6% UR 1yr from now?
(2) Would you have guessed SPY +17% and TLT +5%?
English
JEPA-style models feel especially right for medical imaging. Cooking up in-domain models for x-rays and CTs are so cool- but I’m increasingly convinced that a JEPA-style objective essentially forces a prediction of latent anatomy (rather than just reconstructing pixels)
English
@christyoverflow @elysiuslabs I worked on this idea back at UKDRI, would love to connect!
English
Building a sleep wearable that reads your brain and optimizes sleep in real time @elysiuslabs
If this sounds interesting, would love to connect :)
English
@yieldsearcher I think market is still taking its time to wrap their head around how geopolitics matters more than economics and how this ultimately means data would serve optics from here onwards when needed
English
I 100% agree and for me the head-scratcher is: why isnt the 2Y and SFRZ6 rallying more? (-3bp as we speak)
cc@cchengasaurus
Reading thru the mental gymnastics in the comment section, I think it is clear that regardless of whether this number is fudged that optically and politically the new Fed chair has what he needs to force cuts. White house will run with the Trump defeated Inflation narrative
English
who’s going to make it possible to stay connected to your vibe code session via iPhone
I should be able to interact with my cursor, Claude code CLI, Gemini CLI, Antivgravity, etc. all from my phone
English
This actually costs more than just 1% inflation for most
and is leading to heavy hits on lower decile families
I covered it here, and tbh it's surprising how very few people look into trade compliance data when it comes to macros lol
gingercontrol.com/blog/the-1-700…
Rene Bruentrup@fallacyalarm
The US government currently collects $35bn per month in tariffs and growing. That's 2.5% of monthly domestic consumption. Some of the tariffs may be eaten by foreign and domestic companies via lower margins. But some will manifest as price increases. So we might easily be talking about a 100bps distortion in current inflation data vs prior year. If properly adjusted for this one off effect, the inflation picture might look completely different. Why does nobody talk about it? Why does nobody attempt to quantify it?
English