Welcome back to your daily memorandum talking tech, business, AI, markets, and more. 🗞️

In today’s edition, we are tackling the following:

👥 OpenAI, Anthropic & Google held weeks of AI safety talks.
🦅 NVIDIA's Huang says the AI industry doesn't need new laws.
🤝 U.S. & China discuss tariff cuts ahead of leaders' summit next week.
✨ Meta launches Meta One subscription plans with expanded AI tools.
💣 Trump administration plans $2.8B sale of powerful bombs to Israel.

TECHNOLOGY

• OpenAI, Anthropic & Google held weeks of AI safety talks (CNBC)
  ^{_{More: TechCrunch, Reuters, France 24}}

  • OpenAI policy chief Chris Lehane confirmed weeks of AI safety talks with Anthropic and Google DeepMind.

  • The talks follow Anthropic CEO Dario Amodei's essay urging the industry to slow the development of frontier AI.

  • Some worry the coordination could raise antitrust concerns, though Lehane said the companies don't need a government waiver.

• NVIDIA's Huang says the AI industry doesn't need new laws (Bloomberg)
  ^{_{More: TechCrunch, BBC, Politico}}

  • Huang said AI is just hardware and software built by humans, not an "alien mind."

  • He called safety an engineering problem, not a legal one, arguing existing laws are sufficient for AI oversight.

  • Huang argued that market forces already push companies to avoid releasing unsafe products, so new regulations aren't necessary.

• Mark Zuckerberg sides with NVIDIA's Huang on AI safety & slowdown debate (CNBC)
  ^{_{More: Firstpost, NY Times, Business Times}}

  • Zuckerberg said AI labs face significant liability if their models cause harm, incentivizing safety without regulation.

  • His comments echoed NVIDIA CEO Jensen Huang's market-driven stance, contrasting with Anthropic CEO Dario Amodei's caution.

  • Zuckerberg said that trust and alignment are quickly becoming the most important capabilities that differentiate AI agents and models.

BUSINESS

• U.S. & China discuss tariff cuts ahead of leaders' summit next week (Bloomberg)
  ^{_{More: Business Times, Japan Times}}

  • The U.S. and China are weighing reduced duties on select goods, including American energy and agricultural exports.

  • The talks may signal an extension of the one-year trade truce ahead of next week's leaders' summit.

  • The countries are also expected to lower duties on Chinese manufacturing inputs under an earlier $30B tariff-cut plan.

• Meta launches Meta One subscription plans with expanded AI tools across apps (Reuters)
  ^{_{More: TechCrunch, Yahoo!}}

  • Meta introduced a new subscription service called Meta One, offering expanded AI usage across Facebook, Instagram, and WhatsApp.

  • The plans aim to help Meta monetize its Muse AI models, following its $14.3B investment in Scale AI.

  • Instagram's daily revenue rose 475% to $1.2M, while Facebook's climbed 143% to $528k in the same week.

• U.S. military confirms it deployed weapons into orbit for the first time (BBC)
  ^{_{More: TechCrunch, CNN, The Hill}}

  • Air Force Secretary Troy Meink confirmed the U.S. has deployed on-orbit weapons capable of defending against hostile threats.

  • Officials declined to detail the weapons but said they defend U.S. forces against space-enabled attacks.

  • The U.S. has warned since 2007 about Russian and Chinese space weapons testing, including Russia's nuclear satellite program.

MARKETS

WORLD

• Trump administration plans $2.8B sale of powerful bombs to Israel (Washington Post)
  ^{_{More: Business Standard, NBC, AP}}

  • The Trump administration plans to sell Israel 40k powerful 2k-pound bombs in a $2.8B arms package.

  • The Biden administration had paused delivery of similar bombs two years ago over mass casualty concerns in Gaza.

  • Congress has been informally notified of the deal, mostly funded through U.S. foreign military financing.

• China closes AI gap with U.S. amid growing global safety debate (AP)
  ^{_{More: Economic Times, ABC}}

  • China is rapidly closing the AI gap with the U.S. despite chip export restrictions, per a Stanford report.

  • A Johns Hopkins fellow said the gap between U.S. and Chinese AI models is now narrow and fragile.

  • Both countries have raised AI safety concerns ahead of Xi Jinping's expected visit to Washington this month.

• India to charge merchants fees on larger UPI transactions from October 15 (Reuters)
  ^{_{More: TechCrunch, Bloomberg}}

  • India's UPI will charge merchants a 0.4% fee on transactions above ₹2k ($21), starting October 15.

  • The fee is capped at ₹300 ($3) for larger transactions, and small merchants earning under ₹100k are exempt.

  • Railways, telecom, insurance, and fuel sectors will pay a flat ₹5 fee, ending UPI's zero-fee model that has been in place since 2020.

FUTURISM

• Salesforce & NVIDIA's new Koa model challenges frontier AI labs (TechCrunch)
  ^{_{More: Yahoo!, CNBC}}

  • Salesforce unveiled Koa, its first reasoning model, built on NVIDIA's open-weight Nemotron model at Dreamforce.

  • Koa offers enterprises an open-weight alternative that uses fewer tokens and doesn't ingest customer data.

  • Salesforce previously relied on frontier models such as Claude and ChatGPT for reasoning tasks within its Agentforce platform.

• SpaceX to attempt first Starship orbital flight on September 22 (Reuters)
  ^{_{More: TechCrunch, Inc.}}

  • SpaceX's 14th Starship test flight launches September 22, with a 75-minute window opening at 7:15 a.m. CT.

  • The mission will deploy 26 third-generation Starlink satellites, part of SpaceX's growing 10k-satellite internet network.

  • A successful orbital flight would mark SpaceX's first Starship launch to generate revenue since its June IPO.

• U.S. data centers could surpass Germany & Japan's gas use by 2035 (Bloomberg)
  ^{_{More: TechCrunch}}

  • Data centers are projected to be the second-strongest driver of natural gas demand growth after LNG exports.

  • BloombergNEF forecasts data centers could consume 18B cubic feet of gas daily by 2035, nearly double earlier forecasts.

  • Grid-connected data centers could drive 15B cubic feet of new daily demand, five times other sectors combined.

• Hunting for AI roles eats hours across careers pages, internal referrals, group chats, and a dozen overlapping boards*

  • The Athyna AI Job Board does the watching for you and matches openings to your profile in the background.

  • When a role hits a 75% matching index or higher, we notify you with details and a direct apply link.

CONTENT

• After reinventing technology, Silicon Valley now aims to reshape the human body (a16z)

  • Cyber ethnographer Ruby Justice Thelot joins Elena Burger to discuss wellness culture's blurry line with reality.

  • Thelot explains 'paracontent,' where online conversation about a trend can outgrow the trend itself.

  • They trace wellness trends like peptides, GLP-1s, and wearables, predicting increasingly individualized 'body futurism' ahead.

• Tibo Louis-Lucas runs seven companies with under 10 people using AI agents (This Week in Startups)

  • The Pentagon is reportedly in talks to lend $5B to AI startup Fluidstack, raising equity concerns.

  • Tibo Louis-Lucas explains how he runs seven companies with fewer than 10 people using AI agents.

  • Jason and Lon also mark the 25th anniversary of 9/11 with personal stories during their Off-Duty Friday segment.

• Matteo Franceschetti on obsession, skipping vacations, building a multi-billion-dollar company (20VC)

  • Matteo Franceschetti, Eight Sleep's co-founder and CEO, discussed hiring, angel investing, and lessons on hardware-company growth.

  • He explained how Eight Sleep uses AI to run a $100M email marketing business with no employees.

  • Franceschetti also weighed in on AI's impact on engineering jobs, Anthropic's IPO prospects, and the market potential of humanoid robots.

EXTRAS

• Spotify lets parents exclude kids' music from Wrapped & recommendations (TechCrunch)
  ^{_{More: USA Today, Spotify, Variety}}

  • Spotify introduced a new toggle letting users exclude kids' and family music from their personal taste profile.

  • The feature is rolling out globally on mobile devices for both Free and Premium Spotify users.

  • Parents have long complained that kids' listening habits skew their yearly Wrapped results and recommendations.

• Nitter & XCancel shut down again amid X's legal pressure (Engadget)
  ^{_{More: TechCrunch, Yahoo!}}

  • X sent Nitter a cease-and-desist letter last month over alleged unlawful scraping of its data and API.

  • Nitter's GitHub page was archived on September 11, rendering the open-source project read-only.

  • XCancel suspended its service Monday, citing a new development in ongoing legal proceedings with X.

• Kennedy Center to close for renovations after judge blocks Trump renaming (AP)
  ^{_{More: PBS, Boston Globe, CNBC}}

  • The Trump-aligned Kennedy Center board voted to close most of the venue after a judge blocked the renaming.

  • Congress has allocated $257M for repairs, but Trump said they'd only proceed if his name was added.

  • A federal judge ruled Trump's name cannot be added to the building without approval from Congress.

AND MORE

• Former TikTok execs built an AI app that teaches photo posing.

• South Korea's universities offer a direct route to Samsung & SK Hynix.

• Roko’s Basilisk skips the buzzwords and cuts straight to what moves markets, products, and careers.*

• Wonder scores a $425M DoorDash partnership to expand its food empire.

• Malaysia talks to rival airlines amid concerns over AirAsia's financial health.

• Meta lets AI agents handle boring parts of WhatsApp Business setup.

• Ed Sheeran's tour hits trouble as acts exit in solidarity with Macklemore.

• Here are the AI graveyard projects & startups that didn't survive.