Jing

Someone compressed Andrej Karpathy's entire personal knowledge base workflow into 10 minutes. Obsidian + Claude + your own files, querying a wiki that compounds instead of a chat that forgets. Best AI workflow I've tested this year.

AI “trainers” for finance are getting up to $25k / day to teach AI applications Useful application of AI for pod shops and other analysts = instant model updates based on live earnings call transcripts: “Sinisterra, 30, then walked the class through how to scan transcripts from earnings calls with OpenAI's ChatGPT and Anthropic's Claude to find the most market-moving statements. The machine ran sentiment analysis and translated management’s spoken remarks into numerical spreadsheet inputs to forecast future financials. Participants could see how AI could help streamline some of the most labor-intensive parts of their jobs.” Wild how fast AI is changing the landscape

Yep, we did.

This is WILD! Leopold Aschenbrenner just proved his thesis with his portfolio and the numbers are staggering (Save this). He is a 24 year old former OpenAI researcher who published a 165-page manifesto in 2024 arguing that the real bottleneck on AGI was not algorithms or chips but rather was electricity. He then bet his entire fund on it and the power math he laid out drives every single position. In 2022, the GPT-4 training cluster consumed roughly 10 megawatts and cost around $500 million. AI compute has been scaling at half an order of magnitude per year, so by 2024 the largest cluster was 100 megawatts and cost billions. By 2026, right now the leading training cluster requires a full gigawatt of continuous power, the output of a large nuclear reactor, the power of the Hoover Dam. By 2028, his model projects 10 gigawatts, more electricity than most US states produce in total and by 2030, a single training cluster consuming 100 gigawatts over 20% of everything the United States currently generates costing over a trillion dollars. And that is just the training cluster, inference on top requires multiples of that. Meanwhile, US electricity production has barely grown 5% in a decade and the grid was not built for any of this. His largest position was Bloom Energy, a fuel cell company that generates power directly at data center sites, bypassing the grid entirely. He began accumulating shares in the mid teens through 2025, built the position to $875 million, and watched it grow to approximately $2.73 billion after Q1 2026 earnings showed revenue up 130% year over year and an Oracle deal to deploy 2.8 gigawatts of fuel cells across AI data centers. Bloom Energy stock is now up over 1,400% in the past year. The rest of his Q1 portfolio reads like a master class in the same thesis, CoreWeave at $556 million, Iren Limited at $401 million, Core Scientific at $389 million, and Applied Digital at $320 million, every one of them a power and compute infrastructure play, not a model company. Milk Road Pro called Bloom Energy early before the AI data center power thesis became consensus and our subscribers are up massively on the position. Aschenbrenner's portfolio confirms exactly what we have been tracking, the trillion-dollar AI buildout is an energy and infrastructure trade first, and a software and model trade second. Come join us at the link in bio/below to see our full portfolio before the rest of the market catches on.

🚨 LEOPOLD ASCHENBRENNER IS OFFICIALLY BETTING BILLIONS THAT THE AI HARDWARE BOOM HAS PEAKED. The exOpenAI researcher who was fired for warning that China could steal their AI models then turned $225 million into $5.5 billion in 12 months just filed his Q1 2026 13F with the SEC. One quarter ago he had $5.5 billion in disclosed equity exposure. As of March 31, 2026 that number is $13.67 billion. The portfolio nearly tripled in a single quarter across 42 positions. He initiated $7.46 billion in put options against every major semiconductor company between January 1 and March 31, 2026. None of these positions existed in his Q4 2025 filing. - SMH VanEck Semiconductor ETF PUT: $2.04 billion - Nvidia PUT: $1.57 billion - Oracle PUT: $1.07 billion - Broadcom PUT: $1.01 billion - AMD PUT: $969 million - Micron PUT: $583 million - Taiwan Semiconductor PUT: $535 million - ASML PUT: $494 million - Intel PUT: $159 million For the past 18 months Aschenbrenner was betting only on electricity, memory, compute, and physical data center infrastructure. That made him one of the best performing fund managers in the world. And his long stock book still reflects that exact same thesis. - Bloom Energy: $878 million - SanDisk: $724 million - CoreWeave: $556 million - IREN: $401 million - Core Scientific: $389 million - Applied Digital: $320 million - Riot Platforms: $142 million - CleanSpark: $104 million - Solaris Energy: $62 million - T1 Energy: $43 million - Bitfarms: $38 million - Bitdeer: $29 million - Power Solutions: $26 million - WhiteFiber: $20 million - Babcock and Wilcox: $19 million - SharonAI: $18 million - ProPetro: $13 million - Hive Digital: $6 million He is also running call options on specific names at the same time as his puts, which means he is not simply betting against semiconductors everywhere. - Micron CALL: $422 million - SanDisk CALL: $388 million - Taiwan Semiconductor CALL: $354 million - CoreWeave CALL: $140 million - Bloom Energy CALL: $55 million This means he believes the companies supplying power, storage, and compute to the AI industry still have years of growth ahead of them. But the chip companies that Wall Street has been buying for the past two years at record valuations have already priced in everything good that is going to happen to them. The man who has been right about every major AI trade for the past 18 months is now betting that the biggest names in semiconductors are about to fall. If his track record means anything, the chip stocks Wall Street has been buying for the past two years may be in serious trouble.

leopold aschenbrenner just released his fund's latest investment portfolio & honestly its not what i expected he's gone massively short the entire semiconductor supply chain but also revealed what he thinks the next AI constraint is: > biggest surprise: he's short intel. he sold out of his massive Intel position that made him famous. > he's also gone short $9B worth of puts in nvidia, asml, amd, oracle and van ecks semi etf. > biggest long positions are in memory (sandisk) and power (bloom) with a combined $3.5B value (these are his next bottlenecks) > he trimmed ~$1B of bloom energy position but still holds $1 billion of it. his single largest long holding. > his lumentum position is gone (bearish optics?) > the one consistent thesis is he continues to hold data center bets (ai labs will still need gpus) > fund is now worth $13.7 billion (notional). it was just $5.5B 3 months ago and $220 million ~1yr ago

A big pivot from Ken Griffin on AI: “Number one is, in the last few months, there has been a step change in the productivity of the AI toolkit. It is profoundly more powerful than it was just nine months ago. And for us at Citadel, that has allowed us to unleash a much broader array of use cases for AI. And it has been really interesting to watch, to be blunt, work that we would usually do with people with masters and PhDs in finance over the course of weeks or months being done by AI agents over the course of hours or days. These are not these are not mid-tier white collar jobs. These are like extraordinarily high skilled jobs being, I'm going to pick a word, automated by agentic AI. And I gotta tell you, I went home one Friday actually fairly depressed by this because you could just see how this was going to have such a dramatic impact on society. When you witness it in your own four walls, when you see work that used to be man years of work being done in days or weeks, it's like, wow, like that's the first time I've seen real impact in our four walls.” This echoes my own experience with agents and the conversations I am having with students, friends & clients. The toolkit has dramatically transformed and it feels like in finance, for the first time, AI is real.