Tiangong, a humanoid robot developed by the Beijing Humanoid Robot Innovation Center, posted an 8.86-second time in the large-group 100-meter semifinal at the second World Humanoid Robot Games on Aug. 25, beating the human 100-meter world record by 0.72 seconds. The performance quickly drew attention online, but the race result was only part of the story. After crossing the finish line, the robot failed to slow down, slammed into a protective barrier, fell to the ground, and sparks shot from its waist before it caught fire. The event ran from Aug. 22 to Aug. 26 at Beijing’s National Speed Skating Oval. According to the source article, 666 teams and more than 2,000 humanoid robots took part across 51 competition categories and 1,301 matches. New events this year included weightlifting, tug-of-war, and table tennis, along with first-time two-robot and four-robot simultaneous competition formats. The same report said other marks were also broken during the meet, including a 38.15-second run in the 400 meters and a 2.88-meter standing high jump. It also pointed to repeated failures in boxing, gymnastics, long jump, weightlifting, and dexterous-hand tasks, framing the competition as a display of both rapid progress and major reliability gaps in embodied intelligence systems.

A viral meme titled "The SF Social Contract" has become a sharp shorthand for how money moves through San Francisco’s AI startup economy. Originally created by X user @chiefofautism and later unpacked by BlockTempo, the chart traces venture capital from startup funding rounds into model and API bills, cloud and GPU costs, Bay Area rents, high-cost professional lifestyles, and finally acqui-hire deals by large tech companies. The framing is satirical, but the article argues that the underlying cash loop is real. BlockTempo cites 2026 first-half U.S. venture investment of $412.7 billion, with 86% flowing to AI. OpenAI and Anthropic alone accounted for $217 billion, according to the article. It also says AI startups are spending 40% to 50% of revenue on APIs and inference, sending a large share of operating cash back to the model providers they depend on. Even those providers face heavy infrastructure costs: Anthropic’s annualized compute spending in mid-2026 was estimated at $4 billion to $4.5 billion, or about 60% of annualized revenue, while OpenAI’s ratio was said to exceed 75%. The piece then shifts to Bay Area landlords, citing California’s Proposition 13 and San Francisco rent data from Zumper, before turning to the meme’s dating-and-lifestyle stereotypes and the final exit point in the loop: acqui-hire structures used by Google, Meta, Microsoft, and others to bring in AI founders and research teams without conventional acquisitions.

Demis Hassabis, the DeepMind co-founder who helped lead the AlphaGo project, says a future AI breakthrough on a Millennium Prize-level math problem would be comparable to the system’s famous “Move 37” against Lee Sedol. In his view, reaching that level now appears to be a question of when, not why. The comparison ties today’s math-focused AI race back to the 2016 match in Seoul, where AlphaGo’s unconventional move shocked professional players and changed how many people understood machine intelligence. The article traces how AI systems moved from making basic arithmetic mistakes in 2023 to reaching silver-medal and then gold-medal level performance in the International Mathematical Olympiad through systems such as AlphaProof, AlphaGeometry 2, and Gemini Deep Think. It then follows the shift from contest solving to research assistance and open-problem work, citing cases involving GPT-5, Claude Fable 5, OpenAI’s internal reasoning model, and DeepMind’s Aletheia. Alongside those advances, the piece revisits Lee Sedol’s “Move 78” as a counterpoint, arguing that human value may increasingly lie in framing questions, choosing directions, and working with AI in what Hassabis describes as a scientific “centaur era.”

A roster from the 2000 U.S. Physics Olympiad training camp has resurfaced online, drawing attention because it includes Dario Amodei, now the chief executive of Anthropic. The archived list, preserved by the American Association of Physics Teachers, names 24 students selected for the camp held at the University of Maryland in May 2000. Only five advanced to represent the United States at the 31st International Physics Olympiad in Leicester, where the team finished seventh and all five members won medals. The renewed attention has widened beyond Amodei. The same list also includes Vladimir Novakovski, now founder of Lighter; Badr Albanna, an AI research engineer at Duolingo; and Nilah Monnier Ioannidis, now teaching computational biology at the University of California, Berkeley. The article traces how several of those students later moved into artificial intelligence, quantitative finance, computational biology, computer science and law. For Amodei, the route ran from a physics PhD at Princeton to Baidu’s Silicon Valley AI lab, Google Brain, OpenAI and then Anthropic, which he co-founded in 2021. Others on the roster followed equally distinct paths, and the report argues that the Olympiad network itself became a long-lived talent pipeline whose ties later reappeared in startup formation and venture investment.

A long-running wildlife study at the University of California, Los Angeles has found an unusual funding path: OnlyFans. The UCLA team behind a yellow-bellied marmot research project in Colorado, which has been running since 1962, launched an account called OnlyMarms after traditional research funding became harder to secure. Professor Daniel Blumstein said the page has brought in about $4,000 so far, though the project still needs $75,000 to $100,000 a year to support graduate students and fieldwork. The fundraising effort has also spilled into crypto. Several meme coins inspired by OnlyMarms have appeared on Pump.fun, but Blumstein told NPR that the UCLA researchers did not create the tokens. He said people independently launched a meme coin and suggested the team register it so they could collect transaction fees. The project has also expanded beyond subscriptions, including a partnership with a Colorado brewery on a “Marmot Tears” IPA and a public “Fat Marmot Week” competition. Blumstein said the need to raise money through platforms like OnlyFans shows how far scientific funding in the United States has shifted from the federal grant support that once sustained research like this.

Berkeley dropout-founded AI campus dating service Ditto raised $9.2M in seed funding led by Peak XV Partners with participation from Google's Gradient. Its zero-swipe, no-app model simulates 1,000+ AI dates before matching, achieving a 69% match rate vs. 25% for traditional apps, now serving 42,000 college students.

Citrini, a research firm with a Substack top ranking, caused massive selloffs in U.S. software stocks with its '2028 Global Intelligence Crisis' report. Founder James van Geelen, a UCLA double-major in biology and psychology, transitioned from healthcare to investing, using second-order thinking and a long-term orientation. This article uncovers his cross-industry legend and Citrini's methodology.

An in-depth profile of Citrini and its founder James van Geelen. From the February '2028 Global Intelligence Crisis' report that wiped out hundreds of billions in US stock market value, to the April Hormuz Strait field survey, and the precise call on copper foil in AI, Citrini is emerging as a key bridge between crypto and traditional finance.