Aaron's Picks

中文 · EN

Aaron's Picks shares valuable work I read, watch and listen to. Here I keep my own voice to a minimum and give the original work and its authors the fullest respect. Yet every pick once caught my eye, or opened my mind.

  1. 005 · Podcast

    The Bets That Built NVIDIA

    成就英伟达的那些赌注

    Lex Fridman with Jensen Huang · Lex Fridman Podcast #494

    Ben Gilbert, David Rosenthal · Acquired

    Jensen Huang was a genuinely self-made founder. Through crisis after crisis he pulled NVIDIA back from the brink, built it into the world's most valuable company and carried it to the center of the AI stage. His 33 years at the helm amount to an undisputed business legend, one that deserves to be widely known and long remembered.

    His exit also marks the fall of a role model for ethnic Chinese everywhere, which saddens me deeply. These two sets are my own tribute to what he accomplished over those 33 years.

    This one takes Lex Fridman's interview with Huang as its spine to revisit the technical and business bets he made over NVIDIA's history, decisions that looked baffling at the time and proved visionary in hindsight, from spending a decade in markets that did not yet exist to building a user base first, even at the expense of profit.

    His failings do not cancel out his achievements. NVIDIA's success is something our generation of founders, and many after it, can keep drawing strength from.

  2. 004 · Podcast

    Acquired: The NVIDIA Series

    Acquired 英伟达四部曲

    Ben Gilbert, David Rosenthal · Acquired Podcast, 2022–2023

    Jensen Huang was a genuinely self-made founder. Through crisis after crisis he pulled NVIDIA back from the brink, built it into the world's most valuable company and carried it to the center of the AI stage. His 33 years at the helm amount to an undisputed business legend, one that deserves to be widely known and long remembered.

    His exit also marks the fall of a role model for ethnic Chinese everywhere, which saddens me deeply. These two sets are my own tribute to what he accomplished over those 33 years.

    This one draws on my favorite podcast, Acquired: its three-part history of NVIDIA from 2022 and 2023, plus its interview with Huang himself, tracing the story from a startup idea hatched at a Denny's to the boom that followed ChatGPT. I first listened to all four episodes in 2023, and they gave me a superb foundation for understanding the company and its epic rise over the past three years.

    His failings do not cancel out his achievements. NVIDIA's success is something our generation of founders, and many after it, can keep drawing strength from.

  3. 003 · Speech

    Reading the Success of the Privileged

    优渥阶层的成功解读

    Wei Jiang 姜纬 · Fudan School of Economics, 2016

    This is my favorite commencement speech to date. Wei Jiang, an alumna of my school who was then a professor at Columbia, delivered it at Fudan's School of Economics in the year I arrived. Every graduation season since has brought its parade of prestigious campuses, star speakers and quotable lines, yet this one has never lost its place for me.

    With an economist's cool rationality, she dismantles the success of the privileged and the responsibilities that come with it. It shaped my four years as an undergraduate and several major decisions since.

    In 2022 I was lucky enough to meet Professor Jiang at a Fudan alumni gathering in New York. She was as calm, graceful and luminously rational in person as she is on the page.

    The speech predates the AI era. Rereading it today, I was struck by how far from dated it is; amid all the noise and anxiety, it reads as more clear-eyed than ever. Now that AI is redrawing which abilities are worth paying for, it is all the more worth asking what success is really made of.

    I wanted to share it as widely as I could, in the hope that it lights the way for others as it did for me.

  4. 002 · Report

    GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains

    GDPx:在最值钱的领域衡量前沿模型的能力

    Max Kan, Andrew Megalaa, Dylan Patel · SemiAnalysis

    I have always read SemiAnalysis with a wary eye, and I have little patience for Dylan Patel's arrogance dressed up as objectivity. Personal taste aside, though, this benchmark has real substance and is worth passing on.

    It drops AI into real workplaces. Across the 11 most valuable fields, even the strongest model scores only about 12%, a sign that frontier AI is still a long way from delivering expert work on its own. Every model is strikingly lopsided, and safety guardrails and cost both reshuffle the rankings. There is, in short, no such thing as a best model for every job; the right one depends on the task.

    What makes it interesting is that it follows the money, setting questions only in the 11 fields where AI is likeliest to start paying its way, and grades on results: did the exploit actually land, did the strategy actually make money, does the lab data hold up. It even vets its own questions before setting the exam, and when the same quality check was run on another popular benchmark, it flagged nearly half the tasks. To some extent, that makes this benchmark a more valid measure of true model capability.

  5. 001 · Report

    How employment is changing at Anthropic

    Anthropic 的就业正在如何变化

    Nathan Wilmers, Regina Catipon, Maxim Massenkoff, Peter McCrory · Anthropic Research

    Anthropic's October 2026 report, How employment is changing at Anthropic, was the starting point for APoV 006. After that essay went out, a number of friends wrote to ask where they could find the original, so I turned it into a bilingual summary, with a link to the full report on the final card.