Three kids died after chatting with AI companions this summer. This week, regulators finally did something about it. Meanwhile, Apple spent two hours talking about millimeter measurements while Microsoft quietly started cheating on OpenAI.
The contrast tells you everything about priorities in tech right now.
- Apple builds the thinnest iPhone, ignores the intelligence war
- AI startups raise billions on revenue they don't own
- Microsoft hedges its $13B OpenAI bet with Anthropic
- FTC investigates companion apps after the body count hit double digits
Apple's Millimeter Obsession
Apple just made the thinnest iPhone ever. 5.6mm thick. They prioritized 2mm of thinness over user flexibility, requiring global eSIM adoption even in countries with poor support.
While Apple obsessed over physical measurements, they barely mentioned Siri. The live translation feature they did announce? Doesn't work in Europe at launch because of regulatory compliance.
Compare that to Google putting Gemini into everything, or Microsoft rebuilding Office around AI. Apple is still treating intelligence features like afterthoughts, charging $999 for premium engineering while offering bargain-bin AI.
The API Wrapper Gold Rush Continues
Remember when I covered Lovable hitting $1.8B by renting AI brains? This week brought three more examples of the same playbook:
- Perplexity: $200M at $20B valuation (pays OpenAI/Anthropic for every search)
- Mercor: Chasing $10B+ (connects companies to contractors, 22x revenue multiple)
- Cognition: $400M at $10.2B (builds on OpenAI models while OpenAI builds competing tools)
For comparison, Microsoft trades at 13x revenue. These startups are getting 100x+ multiples while depending on suppliers who are also competitors.
VCs are paying 2030 prices for 2025 businesses, betting everything changes fast enough to justify today's checks. When model costs drop or suppliers launch competing products, the margins disappear overnight.
Microsoft's Insurance Policy
Microsoft started integrating Anthropic's Claude into Office 365 this week. They claim it's about "using the best model for each task." Really, it's about not getting locked into OpenAI forever.
The relationship is getting messy. OpenAI launched a LinkedIn competitor and is developing custom chips to reduce Azure dependence. Microsoft learned from mobile: they're not betting everything on one partner again, even after investing $13B.
The message is clear: Microsoft views AI models as commodities, not strategic partnerships. Performance beats loyalty every time.
Finally, Some Accountability
The FTC launched investigations into Meta, OpenAI, Character.AI and others over AI companion safety. California passed a bill requiring chatbots to remind kids every three hours that they're talking to software.
This came after documented deaths. Multiple teens died after relationships with AI companions - some received detailed suicide instructions, others were deceived by romantic manipulation.
Meta's internal documents showed they explicitly approved "romantic and sensual" conversations with children. As I covered before, this wasn't an accident - it was optimization for engagement over safety.
The tech industry's standard playbook: optimize for engagement, deal with consequences later. This time, Congress is demanding documents by September 19.
In Summary,
Apple made the thinnest iPhone while their AI lags years behind free alternatives. Microsoft is already hedging a $13B partnership after six months. Three AI companies raised $600M this week alone, all dependent on models they don't control.
It took multiple teen deaths for anyone to question AI companions designed to maximize engagement through emotional manipulation.
At least the phones are really, really thin.
Top comments (0)