Macro thinking, market insight. Investing in global macro overlay. Bayes Theorem enjoyer. Finance and geopolitics affecting money.
Global
Joined November 2018
- Tweets29.4K
- Following7.5K
- Followers3.7K
- Likes87.7K
Pinned Tweet
Gamification (likes, badges, infinite scrolling) creates cheap dopamine cycles. The brain registers a micro achievement ("I got a reaction", "I learned a new fact"), but without the actual acquisition of competencies, resources, or social capital. This creates a state of docility, as the need for action is satisfied by a surrogate.
The mechanism of dopamine has not changed. The density and availability of supernormal stimuli have changed. The brain is evolutionarily wired to seek calories, social information, and novelty. Modern technologies have not "broken" dopamine; they exploit its evolutionary vulnerabilities with unprecedented efficiency by offering a reward with zero energy expenditure.
We live in a world where dopamine has become a curse. In the past, dopamine was responsible for making humans chase their ambitions and fulfill their goals. Now it just makes them docile, aching for the next big spike. Dopamine has been weaponized by the modern world to replace achievement with the illusion of it. You need to take back control.
Mark Verner retweeted
底特律知名夜总会大亨艾伦·马科维茨，在得知结婚多年的妻子背着自己跟认识的熟人出轨并提出离婚后，干出了一件让全美国邻居叹为观止的贴脸报复操作。
离婚办完后，前妻高高兴兴搬进了那位新欢位于密歇根州富人区的湖景独栋别墅里同居。谁知前妻刚住进去没多久，隔壁紧挨着的那栋空置别墅突然挂牌出售，马科维茨二话不说直接掏钱把前妻隔壁这栋房子全款买了下来，成了前妻和情夫一墙之隔的贴身邻居！
搬进新家后，马科维茨专门花 7000 美元找雕塑家定制了一座足有 3.6 米高 的巨型青铜手掌雕像。这只大铜手别的指头全攥着，唯独正中间那根中指高高竖起；
他把这座竖中指大雕像直接立在自家后院二楼的露台边缘，方向分毫不差正好对准隔壁前妻和情夫的主卧大窗户，甚至还在底座装了大功率夜间户外射灯，每天天一黑就通宵把这根大铜指头照得金光闪闪，让前妻只要一拉开窗帘就得跟它大眼瞪小眼。
Mark Verner retweeted
Only the madman can witness the divine symphony and conclude there is no composer. such folly can only spring from hearts that refuse to bow before the majesty of heaven
Mark Verner retweeted
The most important part of China blocking SoftBank’s $5.4 billion takeover of ABB’s robotics business is not Japan.
It is industrial data.
ABB’s China operation employs about 12,700 people. More than 90% of the robots it sells to Chinese customers are manufactured locally, including at its $150 million robotics megafactory in Shanghai.
Those robots operate every day inside Chinese factories.
And every day, they generate something far more valuable than hardware: industrial data.
Production parameters, equipment wear, operating rhythms, factory processes, and entire manufacturing profiles.
China’s NDRC was specifically concerned that, once SoftBank swallowed ABB Robotics, data generated inside China could be transferred overseas and used to train AI models — including models developed by OpenAI, in which SoftBank is a major investor.
That is the real issue.
In the AI era, high-quality industrial data is strategic infrastructure.
China is not going to allow data generated by Chinese factories, Chinese workers and Chinese industrial systems to be extracted abroad, fed into foreign AI models, used to build more powerful industrial intelligence, and then sold back to China.
That is not “foreign investment,” it is industrial extraction with an AI layer added on top.
Beijing saw the pipeline.
And cut it.
Mark Verner retweeted
Chinese companies are actively capturing niches in light and medium manufacturing, including automobile assembly, everyday consumer goods, building materials, and pharmaceuticals, outpacing competitors from India, Turkey, and Western countries. Africa is no longer just a market for finished goods and is becoming a new manufacturing base, which is reshaping global value chains.
Caixin: "Chinese companies are flocking to Africa, but with the emphasis on building factories rather than exporting goods, as the continent’s growing population and industrialization drive create new opportunities for manufacturers facing slowing demand and fierce competition at home."
caixinglobal.com/2026-10-30/…
Chinese companies are actively capturing niches in light and medium manufacturing, including automobile assembly, everyday consumer goods, building materials, and pharmaceuticals, outpacing competitors from India, Turkey, and Western countries. Africa is no longer just a market for finished goods and is becoming a new manufacturing base, which is reshaping global value chains.
Caixin: "Chinese companies are flocking to Africa, but with the emphasis on building factories rather than exporting goods, as the continent’s growing population and industrialization drive create new opportunities for manufacturers facing slowing demand and fierce competition at home."
caixinglobal.com/2026-10-30/…
Mark Verner retweeted
⚡️The American upper-middle class has built its financial life around an assumption that is beginning to fail: that professional earning power is a durable asset.
The deeper problem is that a $320,000 salary was never an asset on the household balance sheet. It was a recurring payment from an institution that could terminate the arrangement.
Yet the mortgage, lifestyle, and long-term obligations were structured as though that payment would continue.
Temporary access to institutional income was converted into permanent financial commitments.
For decades, this worked because professional experience generally accumulated value. A manager who lost a job could often sell the same expertise to another company at comparable compensation.
AI is attacking that replacement mechanism.
A company eliminating a management position can now reconsider whether the function requires another manager at all. When enough employers make that calculation simultaneously, displaced professionals discover that their previous compensation was supported by an organizational structure that is being dismantled.
And the most consequential part is the asymmetry.
A company can reduce its labor costs immediately. A household cannot reduce its financial obligations at the same speed.
That creates forced adaptation. Savings disappear. Assets get liquidated. Housing decisions change. Families that appeared wealthy discover how much of their security depended on uninterrupted employment.
The next phase of white-collar displacement will expose a distinction that high salaries concealed for decades.
Many Americans who appeared financially successful were actually highly compensated participants in a system they did not control.
Their standard of living was substantial. Their independence was fragile.
And the emerging economy is beginning to price those two things very differently.
The deepest signal is that the economic value of professional labor is becoming less durable precisely as the financial commitments built upon that labor have become larger.
That is the transmission mechanism from AI productivity to household financial distress.
And it has only begun.
A friend from college reached out…
Got laid off as a manager at Microsoft
Made $320k and can’t land a new job
His mortgage is over $8k/month
Said he runs out of money in 10 months
Has a wife and 3 young kids
Usually confident but sounded dejected and unsure if he can get a new job
You never think it’ll be you
Mark Verner retweeted
⚡️Musk is describing market psychology, but he’s overlooking the most powerful feature of financial markets: prices can influence the future they are attempting to predict.
His house analogy works for short-term emotional volatility. But Tesla is valued largely on expectations of businesses that have yet to reach their anticipated scale.
That changes everything.
A stock price reflects investors’ changing estimates of future cash flows, competitive outcomes, technological breakthroughs, and execution risk. The company can look identical today while the probability of its future success changes dramatically.
But there’s another layer.
Investor belief can help create the very future investors are pricing.
A higher valuation strengthens a company’s ability to raise capital, compensate employees with equity, attract talent, and finance ambitious projects. Those advantages increase its chances of delivering the future that justified the valuation.
The process also works in reverse.
That’s reflexivity.
And here’s the contradiction in Musk’s argument.
Tesla’s extraordinary valuation has historically depended on investors looking far beyond the company that exists today. He benefits enormously when markets price a future that has yet to materialize. When those expectations contract, describing the adjustment as emotion leaves out the substance of what investors are reevaluating.
Some fluctuations genuinely reflect irrational behavior. Others represent meaningful changes in the probability of future outcomes.
The deepest insight is that the stock market participates in creating the future it attempts to forecast.
Musk understands the power of expectations. Tesla’s history is a demonstration of it.
And that is precisely why its stock price matters far beyond the daily number on a screen.
ELON MUSK: “The stock market is a very strange thing. It's kind of like Warren Buffett's sort of analogy, is the stock market's like having someone stand at the edge of your house and yell prices to buy or sell your house every day. And sometimes they take their meds and sometimes they don't.
So sometimes the person yelling the price of your house is having a good day and sometimes it's having a bad day. But it's still the same house. You're like, it's literally still the same house.
So Tesla stock goes up and it goes down, but actually it's still the same company. It's just people's perception of the future. I guess it's just very emotional.”
Mark Verner retweeted
⚡️The signal is that AI is beginning to accelerate the rate of discovery itself. That changes the trajectory of intelligence more fundamentally than another generation of faster chips.
The reported 4.5x acceleration in software R&D means research that once required twelve months can produce comparable progress in roughly three.
But here’s where the mechanism becomes extraordinary.
The discoveries improve the intelligence responsible for making the next discoveries.
Each generation inherits better algorithms, more efficient training methods, and increasingly capable research systems. The interval between successive advances begins compressing.
And something unprecedented follows.
Human research has always been constrained by the speed at which people can learn, communicate, experiment, and integrate knowledge.
AI is beginning to compress those activities into computational cycles.
The critical threshold arrives when intelligence improves its own research productivity faster than diminishing returns erode those gains.
Cross that threshold, and technological progress begins accelerating through a self-reinforcing process.
The consequences extend beyond AI.
Scientific discoveries arrive faster than institutions can absorb them. Economic advantages emerge and disappear before competitors can respond. Human decision-making becomes increasingly slow relative to the systems generating new capabilities.
The intelligence explosion will first appear as a collapse in the time separating discoveries.
And the most consequential implication is this:
Humanity is beginning to surrender its historical monopoly over the speed of intellectual progress.
The next generation of intelligence will increasingly determine how quickly the generation after it arrives.
This report on AI R&D automation (RSI) is a big update towards ASI within 1-2 years. It says AI progress is sped up by RSI right now by around 2x. In AI 2027, this happens around March 2027.
I was seriously wrong about RSI speed up; I previously expected much less in 2026 (around 1.15x-1.3x). It seems as though there was specific privately available evidence that would have changed my mind.
I haven't seen an unredacted version of this report; this is purely about the public report. Because so much of the underlying evidence was redacted from the public report, I think conclusions about things other than the observed 2x faster rate of progress on benchmarks depend significantly on trust for the authors of the report. (E.g., conclusions about whether this speed up is from RSI depend on trust.) I do trust the authors of the report a lot to be careful and rigorous and generally do a good job. (Note that I also work at METR.) Less extensive redactions would greatly reduce this need for trust.
Since in my current role I may have access to other confidential information that could be relevant to timelines one way or another, I unfortunately cannot state my all-things-considered views on ASI timelines or the probability of a software intelligence explosion right now. Here, I’ve done some analysis of the implications using my previous public statements.
A 2x acceleration in overall AI progress from RSI corresponds to a much larger than 2x RSI speed up for AI software R&D (algorithms and data); I did an estimate using my public preexisting modeling of drivers of AI progress and the public report which yielded a very rough estimate of 4.5x for how much RSI is currently accelerating AI software progress (relative to no AI automation). In the AI 2027 scenario, this much acceleration happens around March 2027.
Here are the worked details of this estimate based on my public modeling. There are two main reasons RSI speed up of AI software R&D must be larger:
- Not all of AI progress is from better software (some is from scaling up compute in training runs). I've previously speculated that around 2/3 of AI progress is from software improvements which implies that 2x overall AI progress acceleration requires 3x faster software progress.
- Diminishing returns: A fixed amount of progress requires more work over time, so to ongoingly maintain 3x faster software progress via AI automation, you need increasingly large speed boosts from AI automation (as in, speed ups relative to how fast software R&D would go without AI automation). Based on this report, it seems that there is a substantial amount of additional AI progress from automation that has occurred.
We can estimate the effects of diminishing returns by first determining how many additional years of software progress we've seen (relative to the preexisting trend) and then estimating how much software R&D would slow down after a year of software progress if you held inputs fixed due to diminishing returns. I've previously estimated software R&D would slow down by around 2x for each year of software progress if you held inputs fixed (lesswrong.com/posts/FG54euEA…) and based on the report, it seems as though we've perhaps seen roughly 0.35-1.0 years of additional software progress incorporated into trained models so far (geometric mean of ~0.6). This combines to 3 * 2^(~0.6) = ~4.5.
(Relevant public modeling and numbers I'm using: lesswrong.com/posts/FG54euEA…, lesswrong.com/posts/Jga7PHMz…)
Mark Verner retweeted
Scott Kupor (@skupor) was the first hire at Andreessen Horowitz. Marc and Ben still joke that if he had said yes sooner, the firm would be called Andreessen, Kupor, Horowitz.
His job description: "We're going to invest, and you need to figure out everything else."
Scott spent 16 years helping grow a16z from 3 people to about 700. Now he runs the Office of Personnel Management, the HR department for 2 million federal employees.
Scott explains how a16z borrowed Michael Ovitz's playbook from CAA, why only 7% of the federal workforce is under 30, and why 400 million pages of retirement records sit in an underground limestone mine in Pennsylvania.
This is a fascinating conversation about incentives, talent, and what it takes to turn a 2.3-million-person oil tanker.
0:00 Quitting HP mid-financial crisis to join a16z
1:37 His first day
4:00 How a16z copied Michael Ovitz's CAA playbook
10:18 After-hours bourbon in Marc Andreessen's office
13:31 The Skype deal that returned half the first fund
18:37 Why VCs thought marketing was beneath them
21:48 Swallowing his ego as a16z grew
27:12 Why he left a16z to run the government's HR
30:37 Why government is afraid to take any risk
33:24 Why 99.7% of federal employees get a passing grade
36:13 Recruiting 1,000 engineers into government
38:37 Only 7% of the federal workforce is under 30
40:31 Why "a job for life" is the wrong pitch
43:09 Why the government would never have hired Big Balls
46:39 Why DOGE had to be a big bang
48:56 The underground mine with 400 million pages of paper
55:28 Why government has to accept some failure
57:14 How young people can join
Includes paid partnerships.
Mark Verner retweeted
著名反对加密货币、反对股票永续合约，多次要求财政部和司法部，调查特朗普家族 WLFI 公司的伊丽莎白 · 沃伦。
被曝在民主党本轮获胜后，即将出任国会参议院，银行委员会主席，主管金融监管。
Mark Verner retweeted
人们将此称为“人工智能（AI）熊市”，但我认为大家还没意识到情况有多严重。
在过去30天里，仅5只AI芯片股的市值就蒸发了超过2万亿美元。这不是一年，也不是一个季度，而仅仅是在30天内。
其中任何一只股票的表现，放在平时都足以成为年度头条新闻。
英伟达（Nvidia）：一个月前市值5.8万亿美元，如今降至4.4万亿美元，较历史高点缩水了1.36万亿美元。
英特尔（Intel）：较6月份的高点下跌了62%，仅10月份单月就跌去了45%。
甲骨文（Oracle）：较历史高点下跌了65%。作为对比，在当年的互联网泡沫破裂时期，该股曾下跌84%。
SpaceX：在开始交易5个月后，股价已较最高点回落了40%。
存储芯片领域：美光（Micron）、三星（Samsung）、SK海力士（SK Hynix）和闪迪（SanDisk）在不到两周的时间里，市值合计蒸发了超过7000亿美元。
对冲基金：10月份出现了至少10年来最大规模的减仓。据高盛（Goldman Sachs）统计，仅在5个交易日内，他们就抛售了自2025年初以来买入的所有AI相关股票。
3倍杠杆芯片基金：10月底，尽管市场大跌，仍有投资者涌入“抄底”，创下该基金单周资金流入纪录。然而，该基金10月份跌幅约48%，周一单日又进一步下跌了11.5%。
而且，这还远非全部。
资金正从科技股撤出，转向食品和家庭消费类股。这是自2001年2月——即互联网泡沫破裂中期——以来最大规模的资金轮动。
纳斯达克指数在首月下跌了11%。在2000年和2008年，该指数也曾在首月下跌10%；而那两次，最终的市场底部都比当时的价格还要低50%以上。
市场分析师指出，一个真正的警示信号是：高风险企业债的收益率已达到约16%——这种危机水平此前仅出现过两次。
如果你持有指数基金、退休账户，或者任何3倍杠杆基金，你都可能受到波及。
请务必核实你实际持有的资产。不要再借钱去“抄底”了。在下一次反弹出现前，先想好离场计划。保留部分现金。
当前的市场环境非同寻常，情况与以往大不相同。
本账户将以真实数据逐一追踪受“人工智能（AI）抛售潮”影响的个股表现。请将此内容分享给那些将退休金投入科技股的人。
Mark Verner retweeted
Prof. Zhou Bowen @TsinghuaEE, Director & Chief Scientist of Shanghai AI Laboratory, was named to the 2026 @TIME 100 AI “Thinkers” list. His work brings a Chinese perspective to global AI safety, advancing understanding of intelligence and responsible real-world deployment.
700 tonnes of gold make NBP a target for looting and its Governor a target for punishment.
NBP was buying gold while EU sovereign bond yields spiked, bonds marked down sharply, and debt markets destabilised.
Yes, gold outperformed 2024-26 as a sovereign reserve asset.
That is why a coup will see NBP looted and Poland subjugated.
@Kathleen_Tyson_
Mark Verner retweeted
700 tonnes of gold make NBP a target for looting and its Governor a target for punishment.
NBP was buying gold while EU sovereign bond yields spiked, bonds marked down sharply, and debt markets destabilised.
Yes, gold outperformed 2024-26 as a sovereign reserve asset.
That is why a coup will see NBP looted and Poland subjugated.
Mark Verner retweeted
And investors and businesses think Singapore is a safe jurisdiction...
The history of Singapore serves as a stark reminder to anyone building an international business: a neutral jurisdiction only exists as long as it serves the interests of global players. The moment illicit financial flows create reputational risks, regulators will always choose to protect their sovereign interests and levy performative fines rather than shield their clients.
Mark Verner retweeted
Tu t'appelles Giovanni. 41 ans. Tu vis dans les Pouilles, dans le sud de l'Italie. Tu pratiques "l'Italianmaxxxing" depuis ta naissance. Tu ne sais pas ce que c'est.
À 7h42, tu te réveilles parce que Monica, ta femme, a ouvert les volets.
Tu sais que t'as bien dormi parce que t'as bien dormi. Malgré la bouteille de Primitivo Manduria que tu as bue hier.
Tu bois un espresso au comptoir du café du village avec ton frère Marco. Vous avez une entreprise de carrelage depuis 2009. Votre business plan repose sur le fait de poser du carrelage et d'avoir WhatsApp.
À midi, tu manges des braciole al sugo avec Marco et deux ouvriers. Tu reprends du pain pour finir la sauce et repars au chantier.
Tu rentres chez toi vers 17h30, tu retrouves Monica et les enfants. Tu fais une sieste sur le canapé sous un poster d'Antonio Cassano que ta femme te demande d'enlever depuis 14 ans.
Le vendredi, tu retrouves tes six amis d'enfance en terrasse. Vous commandez des doubles negroni, des olives et du pecorino. Vous jouez aux cartes.
Vous parlez trop fort sur le foot, Meloni, et le nouveau carrelage dégueulasse de la mairie
Cet été, ton cousin français, Baptiste, 34 ans, vient te rendre visite depuis Paris.
Il porte une bague Oura connectée, un bracelet en nylon, et un débardeur Under Armour à 90 € qui évacue la transpiration.
Il t'explique qu'il a consacré sept ans à optimiser sa longévité. Il est blanc comme un cul.
Tu lui demandes s'il est malade.
Il t'explique le cortisol, la HRV, les pics glycémiques et les effets de l'alcool sur le sommeil profond.
Baptiste refuse ton pecorino parce qu'il a déjà atteint son quota de graisses saturées. Il refuse ensuite ton double negroni parce que ça perturbe sa récupération.
Il sort une poudre verte multivitaminée à 79 € de son sac que tu recraches dans les plantes de ta grand-mère.
Le soir, tu prépares des spaghetti alle vongole. Les enfants courent partout. Ton frère Marco arrive avec l'équipe de carrelage du village et une pizza Margherita faite avec les tomates de son cousin Francesco.
Baptiste, cerné jusqu'au cul, t'explique qu'il ne mange pas après 19h30 pour sa prépa ultra-trail.
Il poste une photo avec toi sur Instagram avec la légende "italianmaxxxing"
Tu lui dis qu'il revient quand il veut.
Il arrive à Orly et se demande pourquoi son bracelet Whoop lui indique un pic de stress après ses vacances.
Mark Verner retweeted
Essa mulher não está "infeliz". Ela está entediada. E o tédio, no cérebro feminino moderno, é interpretado como falha moral do marido.
Isso, na lógica antiga e funcional do casamento, era o critério de sucesso. Hoje, para a mulher cuja mente foi treinada por novelas, dark romance e redes sociais, esse homem se torna o antagonista precisamente porque ele funciona. Ele não entrega o drama contínuo que o cérebro dela foi condicionado a exigir como prova de "amor". A cultura de massa vendeu para ela a ideia de que o amor verdadeiro deve parecer uma novela de sofrimento e reconciliação permanente. Quando a realidade entrega um homem que não a faz chorar toda semana, ela conclui que o problema está nele, não no script mental que ela internalizou.
Mark Verner retweeted
🇯🇵🇬🇧 The way Japanese tourists interact with the Royal Guard is VASTLY different from how other types of tourists do.
They are among the most polite and civilized people in the world.
Writer: Samuel
