Macro thinking, market insight. Investing in global macro overlay. Bayes Theorem enjoyer. Finance and geopolitics affecting money.
Global
Joined November 2018
- Tweets29.5K
- Following7.5K
- Followers3.7K
- Likes87.9K
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
apple ne vend plus d’innovation tech depuis plus d’une décennie mais une rente de situation basée sur la captivité de son écosystème, le client paye très cher le confort de la cage dorée ios et un sentiment d’appartenance sociale
Mark Verner retweeted
🚨 شاماث يفجر قنبلة تحليلية عن عقد الحوسبة العملاق بين SpaceX و Anthropic، وكيف يمتلك ماسك "الضربة القاضية" لإعادة صياغة سوق الـ AI بالكامل!
🔥 تفاصيل اللعبة الاستراتيجية الخطيرة:
• ثغرة الـ 90 يوماً: العقد يتيح للطرفين (SpaceX أو Anthropic) الانسحاب وإلغاء اتفاقية الحوسبة كل 90 يوماً بشكل متبادل.
• الضربة القاضية: لو قرر إيلون ماسك إنهاء العقد، فسيفقد Anthropic الجزء الأكبر من قدرته الحاسوبية الأساسية فجأة ودون بديل.
• مضاعفة الأرباح 3 أضعاف: ستتمكن SpaceX من إعادة بيع هذه الحوسبة بسعر السوق الحالي (الذي قد يتضاعف مرتين أو ثلاث مرات)، لتضطر Anthropic للاستسلام الفوري ودفع الأسعار الفلكية!
• تسونامي في السوق: هذه الخطوة ستجبر جميع عمالقة التكنولوجيا وشركات السحابة على الاندفاع لشراء الحوسبة بأي ثمن، وإلا واجهوا الإفلاس المحتوم وخروجهم من اللعبة تماماً.
كل نظرية الألعاب في سوق الذكاء الاصطناعي اليوم تختزل في سؤال واحد: هل سيقوم إيلون بتمزيق العقد؟ لو فعلها.. فالجميع عليه الحذر! ⏱️💥
📌 احفظ التغريدة بمفضلتك لترقب الخطوة القادمة في حرب الحوسبة الكبرى 🔖👇
Mark Verner retweeted
レイ・カーツワイル（世界的未来学者）「2032年、人間は死ななくなる。脳をネットに直結して合体する時代が来る」
寿命が無限に伸びる「不老不死」へ秒読み――体の中に極小ロボットを泳がせ、脳みそを丸ごとクラウド接続！
「ちゃんと健康に気をつけている人は、あと数年で『年を取るスピードより寿命が延びるスピードの方が速い状態』に入って死ななくなります。さらに脳とAIを無線で繋げば、考えた瞬間に答えが頭に浮かぶ超人になります」
天才予言者が明かした「死の克服と脳のAI合体」をわかりやすく👇
・2032年には「実質的に死なない身体」へ突入：医療技術の進化スピードが爆発し「1年生きると寿命が1年以上延びる状態（寿命脱出速度）」へ到達。病気や老化を治すスピードが老化を追い抜くため、健康に気をつけていれば寿命が尽きなくなる計算。
・血管の中を「超極小のお医者さんロボ」が泳ぎ回る：2030年代〜2040年代には目に見えない微小サイズのロボット（ナノボット）が体の中をパトロール。血管の中を泳ぎながら、がん細胞やウイルスを見つけ次第その場で精密にやっつける仕組み。
・スマホを使わず「脳みそをネットに直結」：頭を切り開く手術をすることなくロボットが脳へ入り込み、人間の脳とネット上の超巨大AIをまるでWi-Fiのように無線で接続。
・ググらなくても「最初から知っていた」状態になる：調べ物をしなくても、答えがまるで最初から自分の記憶や知識だったかのように頭の中にパッと浮かぶ究極の超人へ進化。
要は、世界的未来学者のレイ・カーツワイルが「健康に気をつけて2032年まで生き残れば寿命の壁を突破して死ななくなり、血管を泳ぐ極小ロボットと脳直結のAIによって人間は生物の限界を完全に超えた存在へ進化する」と予言したという事です。
Mark Verner retweeted
A lot of gold imports to China are dealer inventories moving to Shanghai and Hong Kong to support 2027 launch of interbank HKPMCC bilateral/physical gold and institutional Project Ensemble digital/physical gold.
Mark Verner retweeted
🇷🇺 Russian gas flows to 🇹🇷 Turkey just jumped almost 4 times.
In the first 5 days of November, Gazprom says deliveries to Turkish customers were almost 4 times what they were over the same days in October and September.
October was already up 78% on September.
Gazprom links it to Europe.
It says Europe pulled record volumes of gas out of storage on 28 and 29 October, and again on 1, 2 and 4 November.
Winter came early for European storage.
Turkey is stocking up on Russian pipeline gas before LNG cargoes get bid away by Europe and Asia.
Gazprom is happy to sell every molecule it can.
And Turkey wants to be the gas hub between Russia, the Caspian and Europe.
Every cold week in Europe makes that hub more useful.
🇹🇷 Turkey wants to power all of 🇨🇾 Cyprus with Turkish gas.
The pipe is 101 km long, mostly under the sea and runs from Anamur to Teknecik in the Turkish Cypriot north.
The plan, as Turkey’s energy minister Bayraktar puts it:
🔹 Turkish gas fuels a gas power plant in the north
🔹 The north could later sell power to the Greek Cypriot south
🔹 The line can run in reverse, so East Med gas could reach Turkey and then Europe
Turkey is the only country that recognises the north, and it doesn’t recognise the government in the south.
Cyprus sits in the middle of the East Med gas field and the Greek Cypriot government is the one with EU backing.
This is a pipeline built as diplomacy.
Whether the south ever buys a single megawatt will decide if it works as an energy project.
Mark Verner retweeted
"Chinese tech companies are getting stronger, evolving into full-stack physical AI generalists, even as they face growing pressure from all sides."
From my new piece on China's tech Swiss Army Knives: highcapacity.org/p/chinas-te…
Mark Verner retweeted
我反而不太认为与股价有关。这个时候放出来这种消息，明显是逼着老黄撤离美国市场，这是为监管英伟达做铺垫。另外，一个中国间谍潜伏在一个科技公司的CEO身边，并不会太多的涉及政治和商业，更像是一种策反行为。共和党不会用这种手法，反而在中期选举，民主党放出消息的嫌疑最大。
Such stories do not emerge in the market by accident. No one knows Jensen Huang's true motives, but seasoned market participants understand this is manufactured noise designed to influence stock prices or create leverage in negotiations with regulators over chip exports. While the crowd focuses on scandals, major players are quietly redistributing assets. Someone is definitely orchestrating this game, and the objective is purely financial.
Mark Verner retweeted
美女率ならモスクワ。
それでもウラジオストクに通いたくなる理由。
ロシア3都市を巡った個人的な印象。
👩美女率：モスクワ≫ウラジオ＞サンクト
🤝フレンドリー度：ウラジオ＞サンクト＞モスクワ
モスクワでは「ミスコンか！」
と思うほどの美女に遭遇する。
でも意外だったのはウラジオストク。
人口60万の港町なのに美女が多く、
素朴で気さくな人が多かった。
3都市それぞれに良さがある。
その中でも、アジアに近いウラジオは、
日本人にも馴染みやすい街だと思う。
Mark Verner retweeted
Uma das descobertas mais libertadoras da psicologia é que quase ninguém está prestando atenção em você.
Em 2000, pesquisadores de Cornell pediram que estudantes entrassem numa sala cheia vestindo uma camiseta constrangedora. Eles calcularam que metade das pessoas tinha reparado. Na verdade, foi menos de um quarto.
Os psicólogos chamaram isso de efeito holofote: você superestima o quanto os outros te observam, porque é o centro do próprio mundo, e acha que é o centro do mundo dos outros.
Pense em quantas coisas você deixou de fazer por medo de uma plateia que nunca existiu.
Os estoicos já sabiam que a opinião alheia não está sob o seu controle. A ciência só descobriu que, na maior parte do tempo, ela nem existe.
Mark Verner retweeted
Assuming an ordinary appearance means someone is harmless is a dangerous misconception. In fact, it is the perfect camouflage.
High emotional intelligence, the ability to be accommodating, and the skill to manage the narrative around the boss are serious professional assets. These people do not just tag along. They shape the agenda. And they always outmaneuver those who underestimate them.
Those who appear to be just a convenient background often hold the most critical levers of power.
Never underestimate those who know how to stay in the shadows while controlling the light.
Mark Verner retweeted
Such stories do not emerge in the market by accident. No one knows Jensen Huang's true motives, but seasoned market participants understand this is manufactured noise designed to influence stock prices or create leverage in negotiations with regulators over chip exports. While the crowd focuses on scandals, major players are quietly redistributing assets. Someone is definitely orchestrating this game, and the objective is purely financial.
Assuming an ordinary appearance means someone is harmless is a dangerous misconception. In fact, it is the perfect camouflage.
High emotional intelligence, the ability to be accommodating, and the skill to manage the narrative around the boss are serious professional assets. These people do not just tag along. They shape the agenda. And they always outmaneuver those who underestimate them.
Those who appear to be just a convenient background often hold the most critical levers of power.
Never underestimate those who know how to stay in the shadows while controlling the light.
Mark Verner retweeted
"These days, nearly every major Chinese tech firm has its own AI models, custom chips, and robots. Instead of EV or robot companies, China now has all-in-one physical AI companies."
From my piece on China's tech Swiss Army knives: highcapacity.org/p/chinas-te…
You are right. Tens of billions are at stake, and Nvidia is not sitting idly by. The company spends millions of dollars on legal lobbying in Washington. Their goal is not to eliminate restrictions entirely, since that is impossible, but to shape regulations that allow them to sell stripped down yet still profitable versions of their chips, exactly as they did with the H20 for China. These are documented financial efforts to protect their market share, and they are far more effective than any anonymous threads on X.
Nvidia possesses absolute information asymmetry. They know their order book, TSMC capacity, and real demand for years to come better than any hedge fund. They do not need to manipulate the market illegally. It is enough for them to provide a slightly more optimistic or, conversely, cautious forecast during an official earnings call at the right moment. This steers algorithmic traders and large funds in the desired direction, neutralizing any Sunday information noise.
A direct short with 10x leverage for Chinese entities is an operational nightmare.
Chinese state or large private funds face extreme regulatory scrutiny from the US (SEC, CFIUS). Obtaining 10x margin leverage on US mega cap stocks through official channels is practically impossible for them. Any such position would be immediately noticed, blocked, or used as grounds for sanctions.
2. Who is actually behind these moves?
Most likely, it is not China as a state, but rather:
• Global macro hedge funds that use the geopolitical narrative (China is squeezing Nvidia) as perfect cover for their speculative positions.
• Offshore proxy structures via Middle Eastern or certain Asian jurisdictions that act in the interests of Chinese capital but mask themselves as independent funds.
• Domestic US players who benefit from driving down prices before a major earnings report or regulatory decision, using the Chinese connection as convenient information noise.
Executing a maximum short with 10x leverage on a stock like NVDA is a kamikaze trade. The risk of a short squeeze (a sharp upward price rebound that liquidates short sellers) is far too high.
Smart money acts differently:
• They buy put options with distant expiration dates. This provides asymmetric risk: maximum loss is limited to the cost of the option, while profit in a crash is unlimited.
• They use statistical arbitrage: shorting NVDA while simultaneously going long on competitors like AMD or the entire sector to neutralize systemic risk.
Chinese tech companies and their affiliated funds benefit from holding a hedge against NVDA because they know new US export restrictions on chips are inevitable. The drop in Nvidia revenue in China is a fact the market has already partially priced in. Information leaks simply help these funds turn a profit on their hedging positions during moments of artificial panic.
In short, thinking there is some Chinese button for a 10x short on Nvidia is an overestimation of direct intervention capabilities and an underestimation of global financial complexity.
The reality is more mundane and cynical: global speculators and algorithmic funds are simply parasitizing the geopolitical narrative.
Mark Verner retweeted
だいぶ明るみになってきた
エヌビディア（NVIDIA）の高性能技術を、なぜ、中国が表立って欲しがらなくなったのか、そして急速に技術的に追い付いてきているのか。
ジェンスン・フアン氏「中国のスパイ」疑われる女性との交際説に…長女の訪韓電撃取りやめ
approach.yahoo.co.jp/r/QUyHC…
Mark Verner retweeted
This is the most jaw-dropping 4 minutes and 21 seconds you will watch this year.
Nicole Shanahan — ex-wife of Google co-founder Sergey Brin, former running mate of RFK Jr., and someone who personally signed nine-figure philanthropy checks — just went full whistleblower on the entire Silicon Valley “tech wife mafia” and how they were used.
Her exact words (full clip attached):
“I don’t think many of the tech mafia wives realize… they were used to set the groundwork for what Klaus Schwab calls The Great Reset.
Their money especially was being conscripted through a network of NGO advisors, Hollywood, Davos, and their own companies.
A really small group of people… completely blind to how their groundwork is being used to enable these Great Reset policies.”
Then she turns the knife inward:
“These women find their meaning through philanthropic work. I really believed I was helping Black communities and indigenous communities rise up.
But now the problems have gotten worse. Crime worse. Mental health worse. The whole model is broken.
At the end of the day they always go: ‘But climate change.’
Social justice + climate change — it gets progressive women 100% of the time.”
She even says many now believe the biggest “climate change issues” are actually geoengineering issues.
This isn’t some random podcast bro.
This is a woman who lived in the mansions, sat on the boards, flew private to Davos parties… and is now saying:
“We were the useful idiots.”
Mark Verner retweeted
🔶If China succeeds in using physical gold as global collateral (incl. REPO) tracked by tokenized blockchain for limited rehypothecation, the process has two phases, accelerated by China’s need to revalue its domestic debt against gold.
⭕️Revalue to equilibrium phase: Existing above-ground gold must first reprice higher so the stock provides enough collateral value relative to financing needs and to clear paper overhangs. China’s large yuan-denominated debt (local governments, SOEs, property) creates a direct incentive to let the yuan gold price rise. This devalues that debt in real terms while increasing the collateral value of the ounces it already holds (official + estimated non-reported). Michael Howell has linked Chinese demand to this internal devaluation dynamic. HK infrastructure (central clearing system, tokenized-deposit DvP targeting commercial 2027, interbank 400-oz bar tokens under Project Ensemble, SGE vault) gives Beijing the mechanism to support a higher equilibrium price without waiting for Western HQLA reclassification.
⭕️Pricing up via usage as collateral phase: Once gold is revalued and functions as verifiable, limited-rehypothecation collateral on a shared ledger, structural demand takes over. Banks, trade-finance desks and collateral managers hold more of it for REPO, margin and liquidity. This demand is additive to jewelry, investment and official buying. Mine supply is inelastic, so the result is a slower but persistent grind higher in USD terms after the initial reprice.
🎯China’s debt revaluation motive compresses the first phase. Ongoing collateral usage then sustains the second. Both phases are now supported by concrete HK moves rather than pure theory.
#Gold #Silver
Mark Verner retweeted
Stanford researchers gave an 8B language model 108× fewer document tokens. It became more accurate at verifying facts.
The project is called Memorilla, an open-source memory system that lets LLMs use large collections of documents through just 16 learned memory tokens.
Instead of repeatedly feeding an entire document collection into the context window, Memorilla encodes documents once and stores their embeddings. When a question arrives, a trainable memory module attends to the collection and produces 16 query-dependent vectors that the frozen LLM can read directly.
The results are particularly interesting.
On the FactKG benchmark, using the same Qwen3-8B model:
→ Full document context: 80.43% accuracy
→ Memorilla with 16 memory tokens: 84.33% accuracy
→ Document tokens passed to the LLM: 108× fewer
Across eight benchmarks covering personalization, factual reasoning, and long-document QA, Memorilla achieved the highest average score among the evaluated methods.
The approach isn't lossless. It can struggle with exact names and fine-grained facts that ordinary retrieval handles better.
But the research demonstrates something worth studying: giving an LLM access to more text doesn't necessarily mean it will make better use of the information.
The code, training recipes, and pretrained checkpoints are publicly available.
Here's the GitHub Repo: github.com/snap-stanford/mem…
