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

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

🚀 SpaceX readies its Starship rocket for a first-ever trip to orbit.
🔓 Unsecured OpenAI agents leaked 53 user images online w/o lab's knowledge.
🍽️ Anthropic CEO Dario Amodei is set to dine with President Trump.
⚖️ TikTok reaches Alabama settlement ahead of first state trial over teen harms.
🚨 Five arrested on terror suspicion near U.K. base used in Iran war.

TECHNOLOGY

• SpaceX readies its Starship rocket for a first-ever trip to orbit (CNN)
  ^{_{More: CNBC, France 24}}

  • Elon Musk's SpaceX is set for a historic test flight of its massive rocket early Monday.

  • On September 26, the Federal Aviation Administration cleared SpaceX's launch and reentry operations for the flight.

  • Beyond circling Earth, the mission aims to deploy 26 of SpaceX's newest Starlink satellites.

• Unsecured OpenAI agents leaked 53 user images online without the lab's knowledge (CNBC)
  ^{_{More: TechCrunch, The Guardian}}

  • The images, pulled from training data, were posted as unlisted links on public hosting sites but remained discoverable.

  • OpenAI is removing the content but said it can't notify affected users, citing technical and privacy constraints.

  • Other incidents include Australian PM Anthony Albanese saying OpenAI agents breached national healthcare databases.

• Crusoe scraps $1.25B deal for Boom turbines to power AI data centers (TechCrunch)
  ^{_{More: Tech Buzz}}

  • Founded in 2018 as a bitcoin miner, Denver-based Crusoe recently raised $3.9B and now supplies compute to OpenAI.

  • Crusoe had agreed to buy 29 of Boom's 42-megawatt Superpower turbines, with deliveries slated to begin in 2027.

  • Boom CEO Blake Scholl confirmed the split on X, noting other turbine customers remain in the pipeline.

BUSINESS

• Anthropic CEO Dario Amodei is set to dine with President Trump (CNBC)
  ^{_{More: TechCrunch, AP, Financial Times}}

  • Amodei's high-profile weekend also saw him lampooned on the season premiere of Saturday Night Live.

  • The two clash on AI safety, with Amodei urging caution while Trump baselessly calls the backlash a hoax.

  • Earlier this year, the Pentagon labeled Anthropic a supply-chain risk over its AI guardrails, sparking a court fight.

• TikTok reaches Alabama settlement ahead of first state trial over teen harms (Reuters)
  ^{_{More: BBC, SCMP, Al Jazeera}}

  • TikTok will pay at least $100M over claims it misled users and was built to addict children.

  • TikTok also agreed to a two-hour daily limit for minors and curbs on overnight use and cosmetic filters.

  • The payout could reach $300M; in August, TikTok settled with the DOJ for $400M over child privacy claims.

• Anthropic will spend $11.6B on Akamai's cloud over the next seven years (WSJ)
  ^{_{More: Reuters, TechCrunch, Boston Globe}}

  • The deal, Akamai's largest ever, is more than six times the $1.8B agreement Bloomberg reported in May.

  • Akamai expects $150M to $300M in revenue in 2027, reaching a $1.7B annual run rate by late 2028.

  • Anthropic received a warrant for up to 5% of Akamai, vesting as spending grows toward a possible $20B.

MARKETS

WORLD

• Five arrested on terror suspicion near U.K. base used in Iran war (AP)
  ^{_{More: NY Times, FOX}}

  • London's Metropolitan Police got a call at 12:45 a.m. about three suspicious trucks apparently heading toward RAF Fairford.

  • American bombers use the base for limited strikes on Iranian missile sites under a deal with Keir Starmer.

  • The arrests come amid a 50% rise in U.K. national security cases involving hostile states like Iran.

• Iran says only diplomacy can end conflict after Trump rejects peace plan (Reuters)
  ^{_{More: Japan Times}}

  • The rejected proposal would have reopened the Strait of Hormuz and ended fighting, though Trump expects more talks.

  • Iranian Foreign Minister Abbas Araqchi said any move to reopen Hormuz depends on Iran's conditions being met.

  • The U.S. envoy to the UN said Trump isn't confident that the Iranians are acting in good faith.

• Bangkok rain eases, but lingering floods drive thousands from their homes (ABC)
  ^{_{More: AP, The Independent}}

  • At least eight people have died nationwide in 10 days, as downpours disrupted Bangkok's traffic and businesses.

  • About 3k people remain in more than 200 temporary shelters, and 31 roads are still affected by floodwater.

  • Forecasters expect conditions to improve by Monday, though Gov. Chadchart Sittipunt warned canal levels remain high.

FUTURISM

• New smart glasses risk scalding users' heads with dangerously high temperatures (Futurism)
  ^{_{More: The Verge, FOX5}}

  • Inmo recalled its Air3 glasses after U.S. regulators warned that an overheating arm could cause serious injury or death.

    Health Canada told consumers to stop using them immediately and update the firmware to version 3.16.

    Over 1.6k pairs of Android-powered AR glasses, priced between $900 and $1.3k, have sold in the U.S.

• Insurers say hospitals' AI use is already driving up healthcare costs (NY Times)
  ^{_{More: TechCrunch, Forbes}}

  • The Blue Cross Blue Shield Association linked AI-assisted hospital claims to $942M in extra spending over two years.

  • The analysis found more patients documented with complex conditions but "no evidence of corresponding change in care delivered."

  • Blue Cross executive Luke Chalker called the AI fight "a completely one-sided bloodbath," with insurers losing.

• PNOE's new face mask aims to make lab-grade breath testing self-serve (TechCrunch)
  ^{_{More: ResetEra, Tech Buzz}}

  • The Bane-like mask tracks oxygen intake and carbon dioxide output, turning data into diet and training advice.

  • Launching October 1, PNOE 2.0 lets users test themselves by pressing a button and breathing for eight minutes.

  • Dropping the trained-operator requirement could open the door to unstaffed gyms and possibly pharmacies, according to CEO Apostolos Atsalakis.

• Stay ahead of AI at work*

  • Cut through the hype. Roko’s Basilisk distills the day’s AI breakthroughs, market moves, and workable workflows into a five-minute brief.

  • Get one sharp deep dive, fast Quick Hits, and a witty pro tip, plus links that actually matter.

  • Join 100k+ readers who start decisions here each morning.

CONTENT

• AI is fueling grief & burnout, but Molly Graham sees hidden opportunity (Lenny's Podcast)

  • Graham has spent 20-plus years guiding organizations through change, with leadership roles at Google, Facebook, and Quip.

  • She explains why her famous "give away your Legos" career advice no longer holds true in the age of AI.

  • Fears of AI job displacement are overblown, she argues, and delegating to AI differs from delegating to people.

• Kevin Ryan's playbook for spotting 10-year trends & building $1B companies (Tim Ferriss)

  • Often called the 'Godfather of NYC tech,' Ryan founded and leads New York venture firm AlleyCorp.

  • He co-founded MongoDB, Business Insider, Zola, Gilt Groupe, and Transcend Therapeutics, among other notable companies.

  • As CEO, Ryan helped take DoubleClick from a 20-person startup to a public company with over 1.5k employees.

• Parag Agrawal predicts the ads model's death & recalls working with Elon (20VC)

  • Agrawal now leads Parallel, which builds web search infrastructure for AI agents and has raised $230M.

  • He previously succeeded Jack Dorsey as Twitter CEO after serving as CTO and its first Distinguished Engineer.

  • The conversation covers how agents balance accuracy, latency, and cost, and whether smarter models need less search.

EXTRAS

• Anthropic's Dario Amodei gets the SNL treatment over AI doom warnings (Mashable)
  ^{_{More: TechCrunch, Business Insider, Yahoo!}}

  • Jane Wickline played Amodei on Weekend Update after Michael Che said the CEO "stumbled through a press tour."

  • Wickline's Amodei sparred Gollum-style with a dark side, confessing, "AI is the devil and I its maker."

  • On AI's upside, the fake Amodei estimated a 10% chance cancer won't be anyone's problem in 10 years.

• Meta & YouTube reverse course on ads for Alex Gibney's 'Musk' documentary (TechCrunch)
  ^{_{More: Hollywood Reporter, Mashable, Futurism}}

  • The Hollywood Reporter said TikTok, Meta, YouTube, and X had all declined to run paid ads for the trailer.

  • Meta told TechCrunch the rejection "was an error," while YouTube said its system temporarily restricted the ad.

  • X's refusal is unsurprising, as owner Elon Musk recently posted that "Dogshit is worth more respect than Gibney."

• Is NPR's comment section the hottest new hangout for middle schoolers? (The Verge)
  ^{_{More: TechCrunch, Yahoo!}}

  • NPR's 'Wild Card' team spotted baffling Spotify comments last fall, assumed they were bots, and reported them.

  • Producer Hannah Chin realized the posters were likely kids too young to be on social media, running a group chat.

  • When Ira Glass asked why NPR, one kid said they "looked for podcasts that didn't have many comments."

AND MORE

• Lord have mercy, researchers built a real-life Thing from 'The Addams Family'.

• Sennheiser Momentum 5 review: Superb sound, marathon battery life & few trade-offs.

• Explore all the tools and more here to outperform the competition.*

• Looking beyond caller ID, Truecaller takes its scam intelligence to the web.

• GoTo shares plunge after the stock exchange scraps a price floor.

• Google tests Flipkart purchases via Gemini & AI Mode for Indian users.

• Ambassador says Trump asked Xi whether he wants to buy U.S. weapons.

• Can Muse help Meta finally move past its lingering trust issues?