X（推特）上的财经、科技与宏观账号
Mistral 4 is pretty good at lawyer bench and cyber security bench, but give it's European heritage, one can only assume it's also good at regulation bench 引用 @MistralAI：Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US
People talk a lot about how the models are extraordinary in verifiable domains (math, coding etc), while not getting much better at things like writing. But is that actually true? Don’t AI-generated stories keep winning prizes and publication in prestigious outlets? 引用 @generativist：the models are
Chevron CEO Mike Wirth on LNG: “It’s not as clean as renewables, and it's not as affordable as coal.”
Shell CEO Wael Sawan on the new conversation around energy: “I think there's a recognition once again that you cannot have national security without energy security. You cannot have an industrial or an economic strategy that is not underpinned with a solid energy strategy.”
Shell CEO Wael Sawan表示，尽管发生了种种情况，中东石油流量目前已接近战前水平的80%以上，显示出许多国家的韧性。
Speaking of higher borrowing costs and vulnerable sectors: This from the front page of today's WSJ.👇 #economy #markets #housing #bonds #yields #mortgages @WSJ https://t.co/wolgoOZkpY
This is from the SA Tokenomics model which is the only granular split I've seen but again this is the 30-40% inference I cite in terms of monetized compute one could make the argument post-training/RL is inference too but if still used for training/internal purposes, I think it 引用 @socialcapital：~
Shanu Mathew称，自2023年以来，达到给定 AI 性能水平的成本平均每季度下降约47%，相当于每年下降13倍。原文称这一降价速度是 DNA 测序的四倍，并附有链接 https://t.co/e73b7e9V7l。
Sites are getting better at blocking agents and it's making them far less effective... very annoying internet needs to be rebuilt and economically structured around agent use most users don't want to deal with the friction of sites, filling out forms, etc. we'll want to speak
Frontier model subscriptions are heavily subsidised. I made a tool to see how much my API cost would have been. It cost me $35 (and counting) to build this tool. I'm on the $100/mo business premium plan. So far I've spent $4853 in code, not counting chats. It's been three https://t.co/XNldSvjdIN
好采访，有时候看看VC的采访更能退后一步看市场和培养耐心 https://t.co/QYs7qcziyq
Don’t these charts suggest that the monthly subs are still just heavily subsidized by the companies compared to enterprise APIs and that’s unsustainable 引用 @kimmonismus：Holy sh*t, Anthropic’s subscriptions offer so much more value, even compared to the new and efficient GPT-6.1 Sol. It’s not even c
It’s this simple folks. Use common sense and the mountain of data we have already on clean energy projects getting held up more 引用 @JaneAFlegal：Land. Solar uses roughly 20x the land of gas per unit of power, wind 100-200x. And we built 322 miles of high-voltage transmission in 2024 vs 530 miles of
Watermarks have limits. They’re often undetectable, especially in short passages. Rewriting or translating text can completely remove the watermark. That’s why, for now, only approved researchers will have access to our detector, so they can help us evaluate and improve the
in case anyone for a second believed this was my inside scoop on what Trump was going to announce https://t.co/1VYRbxcAIY
so we are where we are ... US manufacturing and ag exports are down 2.5% of GDP from their levels in the 1990s ... and are incredibly small v the US economy (~ 4 % of GDP) ... and that decline hasn't actually been offset by growing IP exports (even counting "irish" FDI profits) https://t.co/LZqI4SJo
It also seems obvious to me that "China" together with the dollar's strength (especially v Asia) explains why manufacturing output stopped growing with overall output after 2000 (a break from 70 to 99 ...) (the rising real deficit in goods indicates this wasn't just a shift in https://t.co/uCTcqXPM
A few comments on the China shock debate, from a more "macro" point of view that sets aside the vicious debate about "identification" -- There was clearly a downturn in manufacturing output around the time of China's WTO entry (it doesn't recover til 06/07) 1/ https://t.co/MWnemz6KXK
But it is a myth to say China couldn't supply both the US and the EU market, and thus the US tariffs pushed China into the EU. China has invested at a scale and pace that allows it to comfortably supply both markets ... that it why the Europeans don't have the luxury of time
This chart isn't likely to get better over the next 4 years as interest rates still fell 170 bp from 2016-2020. So perhaps we look at bond in a forward looking way instead of this silly way https://t.co/Gg7iKvHJ1F
“I prefer the world of disruption” - TotalEnergies CEO Patrick Pouyanne (It says a lot about the trading a capabilities of TotalEnergies — and the money it is going to make this year shipping crude via Strait of Hormuz)
Again if you start from the premise that having the most advanced AI is a matter of national security — clearly a view that’s held by many in DC and SF — then the right policy to tame inflation is not to raise rates, but to raise taxes on everything except datacenters. 引用 @colbyLsmith：NEW w/ @benca
This was a really interesting discussion about surging bond yields 引用 @business：On this episode of the Odd Lots podcast, @CarolinPflueger joins @tracyalloway and @TheStalwart to discuss why Treasury and global bond yields have been surging and what central banks have to do to make bonds more “bo
Research Spotlight: AI Utilization and Changes in Economic Performance https://t.co/QUAp77mpuL #SCB https://t.co/XeCJUxUB5q
I agree with @AlexMalitas that the CCC spread widening is a bunch of "problem children" and not a signal of wider credit problems. Too many are over-reading this. Confirming this, @AlexMalitas also noted that Ba spreads (blue repost) are not widening. ---- Which companies have https://t.co/8vfzQjm
would love to see the accounting that includes Chinese made inputs (mechanically, higher prices of Korean made chips and for that matter Canadian oil will reduce China's import share; I am interested in numbers that account directly for embedded Chinese content) 引用 @BaldingsWorld：Let's be clear, t
France &Germany recognize that their trade defenses will always be tested absent a significant adjustment in the RMB -- (they are 100% right that broad rebalancing won't happen without exchange rate adjustment; there isn't a strong link now between Chinese demand growth and https://t.co/ibENlkkL5b
a 1 pp shift in the composition of output would normally lead to a shift of ~ 1.3m jobs (US economy has 130m jobs at the time). the lower estimate reflects the standard view the traded goods sector is more productive than some service sectors, so less job intense
I did not realize it was controversial in some quarters that changes in the overall trade balance lead to changes in the composition of employment and output (and since sectoral employments tends to be concentrated, pronounced regional shifts) 2/2 https://t.co/vvW1Vq6FLS 引用 @Brad_Setser：A few com
A few days ago, I argued that excessive debt wasn't driving yields higher (link). The repost is my argument for why yields are moving higher, and why they are no longer expensive. https://t.co/AcrblgHGGS 引用 @biancoresearch：For years, yields were below where the economy said they should be. That i
For years, yields were below where the economy said they should be. That is no longer the case. https://t.co/m60EDhxYGD
and kind of shocked that there isn't general recognition that the ten years after China entered the WTO were not in aggregate great years for the US economy (jobless recovery, housing bubble fueled by imbalances, global financial crisis, great recession) -- certainly they weren't
Actually I was most interested in the basic observation that a surge in imports had a differentiated regional impact -- which seems to be a very robust conclusion. I had already observed the "China" shock in my own data tracking so that wasn't news to me. 引用 @joefrancis505：My guess is that Dr Set
Wheels were spinning in Berlin, Brussels and Paris over the summer. Beijing didn't do itself any favors by rather flippantly dismissing European concerns even as China's surplus kept on rising 3/3 https://t.co/O340ESNPA9
EPS压制PE的美股，献给还在刻舟求剑，疑惑利率上涨，英伟达没像22年那种下跌却破新高的各位。
Dr Setser citing Autor et al. on the "China shock": "People in those communities experienced bad job market outcomes, bad economic outcomes for many years. So that’s kind of the background. It changed a little bit how many economists think about the cost of trade." Dr Setser https://t.co/pQVmhnlJL
Recent conversations on here have reminded me of this excellent paper: https://t.co/EXOt8Z9fAB "part of the increase in credit from 1999 to 2007 in regions with exposure to trade is a consequence of higher credit demand associated with adverse labor market shocks"
Imports from China started to rise in 21/22 (notably of batteries) -- but on the export side, European auto exports looked resilient going into 2023 ... and then the bottom fell out in the Chinese market (UK's importance to Europe is also obvious in this chart) https://t.co/ao7PTcpXNX
Today, as @SecScottBessent visits Pennsylvania, the contrast is clear: @POTUS' America First agenda is delivering for the Keystone State. https://t.co/4guNB9LMLF
美债收益率越高，未来回报往往越好；自1914年以来，两者关系的R²为0.85。10年前收益率为2%，如今为5.5%；按历史数据，未来十年的年回报约为5%，而非-2%。