The Adjacent Brief

TL;DR: The AI industry's biggest donors are now lobbying for the federal regulation they helped dismantle in 2024, as Chinese competitors erode U.S. model pricing and export controls prove porous. Anthropic's Claude access ban in China has spawned a thriving grey-market resale network, and Asian startups are filling the gap with models explicitly marketed as sanction-free alternatives.

Worth Reading

• Silicon Valley's AI anxiety is burning out founders and VCs alike (paywall) — Bloomberg reports FOMO-driven burnout as a structural condition in the Valley, not a personal failing.
• Apple passing AI development costs to consumers — and Cook says prices are 'unavoidable' — The Verge frames this as a consumer grievance; it's also a test of whether Apple's brand elasticity survives a hardware price hike tied to features people didn't ask for.
• ChatGPT recommended a user for a job — and the user didn't know it was happening — The moment AI becomes a reputational intermediary without consent is worth watching closely.
• Netflix quietly requires a unique email for every profile — Ars Technica covers the enforcement mechanic; Netflix is building per-user identity infrastructure for targeting and churn modeling.
• Europe's refusal to adopt air conditioning is becoming a public health problem — A useful palate cleanser: cultural infrastructure choices have physical consequences, and the cost of ideological consistency eventually lands on bodies.

Culture & Signal

The regulation reversal no one should find surprising

The AI industry spent 2023 and 2024 lobbying hard against federal oversight, helped elect an administration friendly to that position, and is now asking for the rules it previously blocked. The Next Web frames this as irony. It reads more plainly as self-interest at a different stage of market development — early players want rules once their moats are established, because rules freeze the competitive landscape in their favor.

Midterms, data centers, and whose money talks

Bloomberg's look at AI's growing role in the 2026 midterms (paywall) identifies two threads running in opposite directions: industry money flowing into campaigns, and local backlash against data center construction concentrating into a genuine electoral force. Communities dealing with power grid strain, water use, and noise have emerged as the primary source of AI opposition — a constituency far harder to manage with a lobbying budget than Congress.

Sports clips and the broadcast revenue dilemma

CNBC reports that sports broadcasters are caught between distributing short-form clips on YouTube to attract younger viewers and protecting the subscription and linear TV revenue those clips could cannibalize. This tension isn't new — media has run this playbook since the first YouTube highlights went up — but the pressure is sharpening. The broadcasters who built moats around live rights are now watching the clips from those rights become the product that competes with them.

The New Consumer

Algorithm control moves from a buried setting to a feature

Instagram moved its algorithm customization tools out of hidden settings and into the main app menu, per The Next Web. Mosseri's framing is that this is about giving users control. The business logic is simpler: when users feel like they're shaping the feed, engagement goes up and churn attribution moves from "the algorithm is bad" to "I made different choices." The feature doesn't reduce algorithmic influence — it makes users feel responsible for it.

Reputation now runs through models you don't control

A piece in Gettheleverage describes the moment ChatGPT recommended its author for a job — without the author's knowledge or input. Whether a language model surfaces your name when someone asks "who should I hire for X" is becoming a new form of credibility. The implications for personal branding, SEO, and the entire content-for-visibility stack are real, even if the mechanism is still opaque.

Machines & Minds

Enterprise, not AGI, is where the money concentrates

SiliconANGLE makes a pointed case that frontier model vendors are misallocating toward AGI when enterprise AI is where economic value actually accumulates. The argument isn't new, but the context matters: Chinese competitors are compressing Western model margins, which makes the AGI bet look increasingly like a prestige project funded by enterprise subscription revenue. That's a fragile structure.

Anthropic's product direction is agents, not chat

The Creator Economy got inside Anthropic's product roadmap and finds that the company is betting on long-running autonomous agents over prompt-based interactions as its primary direction. Anthropic is building toward Claude instances that run workflows across days or weeks without human input — a different product category with a different pricing model and a different buyer, and that context is the right frame for reading the enterprise AGI story above.

Chinese models close the security gap, and nobody's blocking them

The Wall Street Journal reports that Z.ai's GLM-5.2 matches leading U.S. models at finding security vulnerabilities (paywall), while critics flag that the U.S. has applied minimal friction to Chinese open-weight models circulating domestically. Gary Marcus, writing in his Substack, is blunter: China has largely caught up on frontier model capability, and the gap the export controls were designed to protect is narrower than the policy apparatus assumes.

Export bans create markets they're meant to prevent

TechCrunch documents Asian startups explicitly marketing new models as Anthropic Mythos alternatives that carry no export restrictions. The Wired story on China's grey-market Claude resale networks — "transfer station" sites that buy API tokens abroad and distribute them to Chinese users — shows the demand side of the same equation. Export controls slowed access; they didn't eliminate it. They created a price premium, a distribution layer, and now a competing product ecosystem built specifically to route around them.

Connected World

One person controls more orbital infrastructure than any nation

CuriosityON's piece on Elon Musk's dominance of low Earth orbit puts the scope plainly: Starlink satellites now make up the majority of moving objects visible from Earth at night. The policy question underneath this — what happens when a single individual controls more communications infrastructure than any sovereign government — hasn't found a regulatory home yet. The political friction over data centers in the midterms piece is small compared to the governance gap here.

Matter is four years late and the industry is still showing up

The Verge attended the Unify conference and found Apple, Google, Amazon, and Samsung still actively investing in the Matter smart home interoperability standard, despite the standard's troubled rollout since its 2022 launch. The staying power says something: the manufacturers with the most to lose from a fragmented ecosystem — the ones whose devices don't work together and generate support costs — have a stronger incentive to make a shared standard work than any startup does.

14 articles across 5 themes · 13 sources · Powered by Folo + Claude