Directionally Correct retweeted
Replying to @willdepue @nabeelqu
Directionally Correct retweeted
Nationalize the labs. Zero their equity and you will be known as the greatest President ever, @realDonaldTrump
Directionally Correct retweeted
I was reflecting on my anti-Druckenmiller crusade over the last few days and need to issue an important clarification.
He’s still the GOAT, he just doesn’t have your best interest in mind (he shouldn’t).
While the interview was released last Friday, I went back and noticed it was recorded at the end of January…
Morgan Stanley & Druck obviously knew this, and this gives much greater context to the price action post-publish, and confirms he’s still the GOAT.
(Released at the top in the middle of the day on Friday… they were selling to you)
Well…
That’s going to go down as a historically poorly-timed interview, Druck— let’s review the following two trading days:
1. Korea and Japan bull
Mind numbing drawdowns.
2. USD bear
Most impulsive move up in about a year.
3. Copper bull
-15% drawdown from ATH close on the day
4. Very little use available from TA today
I’d argue it’s never been more relevant… Very basic forms of it is what allowed me to sell the top of the Kospi at 150 bucks… the irony.
Directionally Correct retweeted
He was real as fuck for this. I hate capeshitter journalists going "bro it's just le hecking sci-fi" who refuse to engage with the technology on its own terms.
《英伟达之芯》的作者,在采访黄仁勋时,就被黄狠狠训了一顿,难怪他心有余悸,印象深刻。
他在这本书的最后一章,记录了黄的这次发飙:
不知该如何继续话题,我决定向黄仁勋展示阿瑟·克拉克的一段文字,就是这段文字激发了布莱恩·卡坦扎罗热情奔放的演绎。在克拉克滔滔不绝地描述机械进化的前景时,我注意到黄仁勋的脸色渐渐发白,好似面部的血色都被抽走了。
“我觉得你找错采访对象了,”他低声嘀咕,“你现在仿佛在采访埃隆·马斯克,可我不是他。”我愣了一下,不知所措,但反应过来时已经晚了。我显然触碰到了他的敏感点。
“机械进化会……会毁掉工作机会吗?”黄仁勋越来越气愤,声音也不由自主地提高。“难道计算器的出现就毁灭了数学吗?这种问题真是老掉牙了,我听得耳朵都生茧了,”他抱怨道,“我不想再谈这个话题了!同样的问题,反复讨论,没完没了。想想看,我们发明了农业,让生产食物的边际成本几乎归零,这对社会难道不是好事吗?我们大规模生产电力,使砍伐树木、生火、携带火把的成本几乎为零,然后我们就可以腾出手去做更多其他的事情了。紧接着,我们会去降低计算的边际成本——长除法!我们将其降至零!”他的嗓门开始提高:“我们一直在努力将各种事物的边际成本降为零,一代一代又一代,而这样的争论每次都会上演!”
我尝试转移话题,但未能如愿。他的话语中流露出愤怒与厌恶。他开始以对待叛逆少年的口吻训斥我。他坦言,曾对我抱有厚望,如今却大失所望。我耗费了他的时间,也浪费了大家的时间;现在,整本书的策划都受到了质疑。当时采访现场有黄仁勋的两位公关代表,但他们均未尝试介入——显然不想引火烧身。
柯克猜测,黄仁勋的愤怒可能是策略性的表现。但说实话,我当时并未感受到这一点。他的愤怒显得肆无忌惮、铺天盖地,甚至太不合时宜了。我并非黄仁勋的员工,他对我发火并无任何益处。针对自己所创建工具的负面问题,他简直烦透了。他认为这个问题十分愚蠢,且已被反复问及太多次了。
“这绝非荒谬的科幻故事,”他说着,指向桌子另一头两位目瞪口呆的公关代表,“你们懂吗?我并非在诸多科幻故事的熏陶下长大,这可不是什么科幻片。这些人在做的是严肃的工作!”他继续说道:“这不是闹着玩的!这跟克拉克毫无关系,我压根儿没读过他的破书,也毫不在意那些书!这不是——我们不是在重演科幻剧!这家公司绝非《星际迷航》的翻版!我们做的可不是那种事!我们是认真的人,做认真的工作。再者,这是一家认真的公司,我是个认真的人,只是做着认真的工作。”
接下来的20分钟,黄仁勋以责备、恼怒与轻蔑交替的语气,质疑我的专业精神、采访方式,以及对项目的投入程度。他指责我企图对他进行精神分析,表示他多么反感回答我的传记作者式问题,尤其是那些试图挖掘他心理状态的问题。“我讨厌这些刨根问底的问题,”他说,“我不喜欢谈论自己,好吗?!我不是在做心理治疗。”他暗示我的问题很愚蠢,形容它们“乏味至极”。他否认自己有任何特别之处,尽管有悖所有的证据。“听着,我——我极其普通。”他说。
“我从未见过像你这样的人。”我说。
“我极其普通。”黄仁勋说。
慢慢地,愤怒平息了。黄仁勋数次转移话题,聊起即将推出的产品,询问他的公关代表状况,并缅怀已故的约翰·尼科尔斯对CUDA项目发挥的重要作用。在某个时刻,他谈起了罗马帝国。接着,他又轻声地责备起我来,显然,他对我已经失去了兴趣。于是,我被匆匆地请出了门。离开采访现场时,我感到十分困惑。虽然我与高管们有过不少紧张的对话,但从未有人像这样对我发火。我深感震惊——不过,说实话,我也感到一丝莫名的兴奋。能成为黄仁勋愤怒的对象,从某种意义上说,竟也算是一种荣幸:这是每个进入他核心圈子的人都必须经历的考验。走出会议室时,我转向其中一位公关人员。
“这次还挺有意思的。”我说道。
他笑了笑,回应道:“哦,是吗?这还不算什么呢。”
Directionally Correct retweeted
TEXAS GOVERNOR-ELECT GINA HINOJOSA: DON'T HATE DATA CENTERS, HATE HOW THEY ARE TAKING ADVANTAGE OF TEXANS-CNN INTERVIEW
this genuinely topped the market
WTF is a SOTA Wife?
SOTA Wives is a podcast exploring the unfiltered side of Silicon Valley.
We built it as an antidote to the performative polish of tech Twitter.
Brought to you by @MilksandMatcha x @vivianmshen.
Episode 1 goes live on Monday, 12th October at 9 AM PT on everywhere you get your podcasts 💙
Directionally Correct retweeted
Replying to @DuckPond86
You make the most money buying high beta garbage in uncertain times
Replying to @___4o____ @SOTAwives
Directionally Correct retweeted
Replying to @valuedrift
yeah ramaswarmy losing ohio / dems winning will weigh negatively on $ASST in addition to general crypto beta
Directionally Correct retweeted
Boomer wealth is going to be extracted from them entirely by the healthcare industry and their families will get nothing.
My coworker's mother passed away last November.
She left her modest $190,000 house to him clear and free, no mortgage, completely paid off after thirty years of factory work.
He moved into the home in January to finally settle the estate.
Two weeks ago, a letter arrived from the state Medicaid estate recovery office.
Because his mother spent four months in a rehabilitation facility before she passed, the state quietly slapped a retroactive $114,000 lien against the property title.
He called the state health department, frantic, asking how a hospital stay could bankrupt an inherited home that had no mortgage balance.
The caseworker sighed and told him:
"Federal law requires state programs to claw back healthcare costs from deceased estates. The house cannot be sold or transferred until the balance is cleared in full."
He offered to pay a monthly settlement.
They said no. The full $114,000 is due in 60 days, or they foreclose on the property at a public auction.
A lifetime of working-class savings, wiped out overnight by a retroactive healthcare lien.
You don't own generational wealth.
You're just holding a property title until the state finds a legal mechanism to repossess it the moment you die.
Directionally Correct retweeted
完蛋了…..以后中国开源模型只能开源小参数的模型了….
商务部科技部就修订《中国禁止出口限制出口技术目录》征求意见,为期一个月,涉及人工智能(大模型、智能体、算力集群、信息处理等23项)生物医药、光伏3个领域。
1️⃣限制出口AI大模型:超过10^26计算量、或2T参数、或公开评估排名全球前10的大模型。
2️⃣限制出口智能体Agent:工具调用与系统操作技术,支持集成的第三方工具数量≥200个,工具选择准确率≥90%,且在需要调用3种及以上工具协同完成任务场景中,端到端任务平均成功率≥90%;自主规划与任务分解技术,可将自然语言指令自动拆解≥20步可执行子任务链,子任务分解精度≥80%,且子任务平均成功率≥90%;标准博弈基准超过90%的人类的博弈决策技术;得分75%以上的AI辅助漏洞挖掘技术。
限制出口算力集群技术:片间互联技术、分布式训练容错技术、通信调度技术、异构集群高效推理与调度。
征求意见期间,出口相关技术需向商务部填报登记表。已通过开源、发表学术论文等方式进入公共领域的技术,不纳入管理范围。
(中国商务部)
目前看来@Kimi_Moonshot 的K4、@MiniMax_AI 的M3 Pro、@Zai_org 的GLM5以及@deepseek_ai 的下一代模型都没法开源了…..
而且相关架构、训练优化和推理加速技术也都不行了。
Readers added context they thought people might want to know
该帖称中国大模型今后“没法开源了”、只能开源小参数模型,与官方文件不符。公告明确支持人工智能开源创新生态,对已通过开源等方式进入公共领域的技术不纳入管理范围,并对开源情形适用自动许可等便利措施(仍在征求意见中)。
mofcom.gov.cn/zwgk/zcfb/art/…
mofcom.gov.cn/syxwfb/art/202…
Directionally Correct retweeted
Some observations on 'autoresearch', in our experience:
1) 'make this thing faster': work wells.
2) 'improve this frontier training run' deeply unimpressive.
3) 'i want to catch up to near frontier' works fine.
(2) is obviously the dream, basically = RSI. We don't have that yet, clearly.
(1) can at times be (2), although generally its not been for us, so it works (for code not at frontier already). Plus, 'code' easier than 'research', at least easier to posttrain for, so, I think it follows. I do hate drawing the distinction with sharpness though.
(3) is pretty useful - straight labor substitution, you could do it yourself since its basically running a playbook, but don't have to.
Agents can be helpful for (2) type work too of course, experiment manager on steriods.
I guess the quadrillion dollar question is when we cross the line for (2). I think soon enough, although one can make an argument that RLVR is not a great paradigm to get there (but I wouldn't make a big bet against it being enough).
Directionally Correct retweeted
Market is now just a guessing game of which 10 tech stocks get gamma squeezed at the open
Directionally Correct retweeted
my read on why ant pushed back the ipo is that they have a lot of promising results in silico and they are still waiting on some preliminary verifications from CROs. an imminent exciting biotech result could also be why situational awareness is playing short term call options
Directionally Correct retweeted
Replying to @goodalexander
alright look
i’ll be the first one to admit that the benefits of ai aren’t felt nearly as much as the hype would suggest
but you’re exaggerating
1. it’s not just writing code. accounting is much easier, legal stuff is much easier, professional photography is much easier (editing photos used to take weeks, now takes seconds), commerce is easier, lots of professions are adopting this
2. gta 6 is the last mega-game ever. there won’t be more. meanwhile micro-games are printing on roblox and at least 50% of the most profitable ones are heavily leveraging ai (if not complete and utter slop)
3. yeah the ai influencers are cringe. no notes there
4. claude killed khaminei. but you’re not ready for that conversation
5. muse will have 500 million users soon
6. higher
Directionally Correct retweeted
“Hi, I believe there’s a 90% chance that AI will wipe out all life on earth.
What am I going to do about this, you ask?
Why of course, I plan to spend my last days creating eval datasets at a non-profit organization in Berkeley, CA!”
Directionally Correct retweeted
what you wanted: Robotic maids, working sanitation systems, on-time flights, elimination of rote administrative jobs, decline in healthcare costs
what you get: increasingly elaborate fucking LARP about civilizational AI running in an unverifiable sandbox to pump an IPO
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
I’ve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
dwarkesh.com/p/openai-huggin…
Directionally Correct retweeted
In two years from now you will be able to buy a neocloud for less than they raise at today. Just saying.
Directionally Correct retweeted
The only crime you can be investigated for in Trump’s America is losing money
BREAKING:
The SEC has launched an investigation into the trading of Leopold Aschenbrenner's Situational Awareness fund, per NYT.
The SEC sent subpoenas to multiple banks that gave Situational Awareness money to trade with, asking for details about the timing of trades and how Situational Awareness communicated with lenders.





































