Jeff Greenfeld

1/16 the ai industry is making an unusual bet: buy compute for years, sell intelligence by the token. banks, commodity merchants, airlines, utilities and cloud providers each explain a different part of the risk.

Funny, after deploying billions I've rarely invested in a founder who didn't start with a more junior partner ...

SpaceX Blocked A U.S. And NATO Satellite to Assist a Russian Intercept After Crippling Ukraine's Air Defenses The Math Proves the Reality that Musk doesn't want you to see. On Oct 13-14, SpaceX restricted Starshield/Starlink during a 39-missile Russian attack, cutting Ukrainian drone intercepts by 23% and stranding U.S. personnel in Kyiv. Forty-eight hours later (Oct 15), SpaceX blocked U.S.-licensed NRO, Space Force, and NATO radar satellite ICEYE-X36 in orbit so Russia's COSMOS 2612 weapon could intercept it (99.89%-99.98% posterior) Our Model Demonstrates that the Russian intercept of the US/NATO ICEYE X-36 satellite (contemporaneous with the throttling of Starlink) would not have been possible, and almost certainly not been attempted, without SpaceX resources. At a material opportunity cost, Russia sprinted its COSMOS satellite to intercept ICEYE X36. Based on a previous attempt in June during which X36 maneuvered to evade COSMOS, Russia knew that this would be successful only if X36 could not maneuver. On October 15, SpaceX fixed X36 in place by issuing a collision alert for a sub-kilometer approach with its Starlink 35856 when SpaceX's own data shows that the actual distance was 4.8-5.0x as great. SpaceX issued no warning to COSMOS despite the fact that its sprint was far more threatening than the stable X36 orbit. SpaceX did not maneuver its Starlink 35856 satellite to clear the supposed collision risk. How, absent coordination with SpaceX, could Russia have known that SpaceX would take actions that (a) misrepresented the danger posed by X36, and (b) resulted in the fixing of X36 thereby enabling COSMOS's successful intercept? Thus, SpaceX was directly involved in simultaneously compromising Ukraine’s defensive and offensive capabilities. The conjunction of these two events makes innocent explanations untenable. The Pentagon Capture & Iran Round Two Trap: Pentagon CTO Emil Michael who cashed out $7M–$35M from Musk’s inner-circle SpaceX vehicles (M8/Gigafund 0.1 and M6/Valor M33 II) and KQ Partners / xAI (selling xAIfor $5M–$25M three days before taking over CDAO/OSCand 21 days before SpaceX bought xAI, while keeping M8, M6, M9, M10 & Nomadic open) and Pete Hegseth purged rival Anthropic in two hours, surrendered combat command over US drones used to attack Iran to Starlink’s "Terms of Service," and on October 2 named Musk Co-Director of Project Meridian to design the Pentagon's "Future of Warfighting" just as America enters Iran Round Twowith depleted Patriot/THAAD magazines and Russia feeding U.S. warship/aircraft locations to Tehran. SpaceX’s actions in Ukraine raise serious questions about its alignment with US security interests and its reliability as a cornerstone of US defense. These questions are particularly troubling given the effectively irreplaceable position SpaceX has been granted by the Hegseth/Emil Michael Pentagon. $SPCX $TSLA Our full quantitative analysis below w/ @streetwiseprof

must feel good to be an Anthropic co-founder > help build chatgpt at OAI by training on world's collective knowledge > keep model weights closed to extract max profit from every token you sell > leave and launch a direct competitor > spend years and millions lobbying washington on how dangerous your own models are > get designated a supply chain risk by the pentagon and fail to ban OSS anyway > pay $1.5bn to settle over 500k pirated books (while watching equity value rocket past $2tn) > predict you will soon create a "country of geniuses in a data center", warn that half of entry level white collar jobs will vanish, and sell the dream of curing cancer > have your staff call it presumptuous to expect your company to make money > meanwhile have your company buy back your shares above fair value, then sell again in a second tender and into the IPO > pledge 80% to charity, which your own s1 calls "voluntary, revocable" and "not legally binding" > keep a fixed 50.1% of the votes through one share, even after selling 98% > help pick the trustees who pick the board > park 5% of the company in a "public benefit fund" you co-run > have shareholders cover the private jet for personal use > call it a public benefit corporation anyway