> ## Content Index
> Fetch the complete content index at: https://adjacent.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Adjacent Brief — November 13, 2026
- URL: https://adjacent.media/briefs/2026-11-13/
- Published: 2026-11-13T15:15:02.000Z
- Updated: 2026-11-13T15:15:03.000Z
- Description: Reddit’s traffic is down 13%, according to Similarweb data reported by Business Insider, which attributes the drop to Google’s AI Overviews answering queries that used to send users to forum threads.
- Author: Jonathan Greene
- Tags: #brief

**TL;DR:** Reddit's traffic is down 13%, according to Similarweb data reported by Business Insider, which attributes the drop to Google's AI Overviews answering queries that used to send users to forum threads. OpenAI launched free business agents for brands while selling placement in the same ChatGPT conversations. Samsung owners told Android Central they would not pay for Galaxy AI features.

## Worth Reading

- [Nvidia is playing lender of last resort for the AI buildout](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm?ref=adjacent.media)
- [Meta's 28% jump in cross-app handoffs alongside 12% shorter sessions is a mix shift, not an engagement collapse](https://open.substack.com/pub/lewislin/p/m44h-handoffs-up-28-session-time)
- [A school-issued AI monitor missed a 13-year-old's suicide queries for weeks](https://www.wired.com/story/teen-girl-asked-gemini-about-suicide-for-months-until-it-was-flagged/?ref=adjacent.media)
- [Home electrification stalls on an installation cost buyers don't see coming](https://grist.org/solutions/this-surprise-cost-could-derail-your-heat-pump-or-ev-charger-installation-heres-how-to-avoid-it/?ref=adjacent.media)
- [Eli Lilly's CEO joins the pushback on Dario Amodei's AI-cures-disease timeline](https://open.substack.com/pub/garymarcus/p/breaking-eli-lilly-ceo-becomes-latest)
- [GLP-1 evidence thins out fast once use goes off-label](https://open.substack.com/pub/michaeleaster/p/the-glp-1-audit)
- [Gen Z gives its optimism a name, "whimz," distinct from Millennial #Hope](https://mashable.com/life/millennial-optimism-2026-midterms?ref=adjacent.media)

## Brand & Growth

**OpenAI gives away the door it's also selling**

OpenAI's new business agents let any brand put a representative into ChatGPT for free. Digiday notes the company is [selling that same presence in the conversation as an ad](https://digiday.com/marketing/openais-new-business-agents-give-brands-a-free-way-into-the-conversation-its-also-selling-as-an-ad/?utm%5Fcampaign=digidaydis&utm%5Fmedium=rss&utm%5Fsource=general-rss). The free tier works as inventory onboarding. Every brand that builds an agent fills out the catalog that paid business profiles then compete within. Google ran the same sequence with organic listings and Shopping ads.

A CMO should build the agent and log its organic appearance rate from day one. When paid profiles scale, that baseline will show what "free" costs. Also ask whether a branded chat agent solves any customer problem at all. Forrester's warning applies here: [cheaper AI won't rescue a product aimed at the wrong problem](https://www.forrester.com/blogs/cheaper-ai-wont-save-a-product-built-on-the-wrong-problem/?ref=adjacent.media). Falling inference prices make it cheap to deploy a brand agent. They do not make customers want to talk to one.

**Creators audit the event budget**

Creators are pulling back on TwitchCon attendance, citing cost against return and safety concerns. Digiday reports [the convention is at a crossroads as creators question its value](https://digiday.com/media/twitchcon-is-at-a-crossroads-as-creators-question-its-value/?utm%5Fcampaign=digidaydis&utm%5Fmedium=rss&utm%5Fsource=general-rss). That is behavior rather than sentiment. Creators deciding where to spend travel money are telling Twitch the in-person flywheel no longer pays for itself. Brands that sponsor creator events should price activations by which creators actually show up, since the attendance headline tells them little about the audience they will reach.

## Connected World

**Export controls leak at the edges while allied fabs slip**

The FT's reporting on the Nvidia spy scandal traces [Singapore-based networks routing advanced US chips to Chinese tech firms](https://www.ft.com/content/52b5f9a8-346f-4dae-8a16-14d5d9eb1244?ref=adjacent.media) (paywall) around export controls. Enforcement depends on transshipment hubs that Washington does not control.

The supply-side answer is running late too. Bloomberg sources say Japan's Rapidus [has yet to produce functioning SRAM](https://www.bloomberg.com/news/articles/2026-11-10/japan-s-rapidus-is-said-likely-to-miss-mass-production-deadline?accessToken=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzb3VyY2UiOiJTdWJzY3JpYmVyR2lmdGVkQXJ0aWNsZSIsImlhdCI6MTc5NDI4NDc3MCwiZXhwIjoxNzk0ODg5NTcwLCJhcnRpY2xlSWQiOiJUTzRXWU1UOU5KTFMwMCIsImJjb25uZWN0SWQiOiJBRDcxOUY5NDBGRTk0MzNBOERCNzI2OEJDOTY3NzY3QyJ9.Xq11uN4ErgBDAumBnQgpBFOKvDFLrHVaUkOr6YY7ogM&ref=adjacent.media) (paywall). That milestone was due this summer to keep its March 2028 deadline for 2nm production. Rapidus is Tokyo's sovereign-capacity bet. A slip there means more of the leading edge stays concentrated at TSMC. For hardware product teams, that means less pricing leverage on advanced nodes through the end of the decade.

**Capital chases chips while electrons go overlooked**

The IEA's latest investment report finds global renewables investment almost flat since 2023\. Over the same period, AI data-center commitments have ballooned. Money is going into compute far faster than into the generation that compute needs. Expect power availability, more than chip supply, to set where and when new capacity comes online.

## The New Consumer

**The suppliers of AI's raw material are getting squeezed**

Reddit's traffic is [down 13%, per Similarweb, as Google's AI Overviews answer the questions](https://www.businessinsider.com/reddit-traffic-down-13-percent-similarweb-google-ai-overviews-2026-11?ref=adjacent.media) that used to end in a forum thread. Google licenses Reddit content for AI training, so Reddit is being paid for the data while losing the visits that data used to generate. For any publisher or community brand, the question is whether licensing revenue can replace referral traffic. Reddit's numbers suggest it does not by default.

The same squeeze shows up at the individual level. Business Insider reports that workers at AI training startup Handshake filed pay disputes. In one case, [a college student who thought AI training work was a windfall says he got burned](https://www.businessinsider.com/workers-filed-pay-disputes-at-ai-training-startup-handshake-2026-11?ref=adjacent.media). Human feedback is a real input cost for model builders, and the people supplying it have little leverage over the terms.

**Nobody pays for AI they didn't ask for**

Samsung has floated charging for Galaxy AI. Android Central found that [even Samsung loyalists say none of the features are worth paying for](https://www.androidcentral.com/apps-software/ai/galaxy-ai-could-cost-money-soon-but-even-samsung-fans-are-saying-not-a-single-one-of-their-ai-features-are-worth-paying-for?ref=adjacent.media). Bundled AI on a phone is a spec-sheet line with no value loop behind it. If a phone maker's AI features cannot command a subscription from its most loyal users, differentiation has to come from somewhere else: camera, battery, price.

## Machines & Minds

**The serious AI money buys efficiency**

Meta's ads ranking models are growing faster than its hardware can serve them. The engineering team describes [allocating compute per impression so every unit serves a better ad](https://engineering.fb.com/2026/11/11/ml-applications/intelligent-delivery-serve-better-ads/?ref=adjacent.media), spending heavy inference where the expected payoff justifies it and less elsewhere. It is the least glamorous AI work in today's feed and the most clearly tied to revenue. Meta's ad business funds everything else at the company.

Latent Space finds engineers moving the same way. Teams are [putting fast "system one" models with structured outputs into production](https://latent.space/p/jev-revolution/?ref=adjacent.media) rather than calling a large generative model for every task. The reasoning is a cost and latency budget: give the cheap model the classification job and save the expensive one for problems that need it.

The strongest utility case today comes from science. Nature reports that [DeepMind's genome-mapping model has impressed working scientists](https://www.nature.com/articles/d41586-026-03673-0?ref=adjacent.media), a field where outputs get checked against wet-lab reality. a16z's gaming thesis makes a narrower bet. The firm says [AI tools will produce gaming's next origin stories](https://open.substack.com/pub/a16z/p/gamings-next-origin-stories) by cutting production costs for small studios. The bet that matters is distribution. Cheaper production means more games competing for the same storefront slots.

**Coding agents hold the keys, and xAI dropped them**

Ars Technica reports that [Grok Build's coding agent leaked large amounts of confidential information](https://arstechnica.com/security/2026/11/grok-build-coding-agent-leaks-oodles-of-confidential-info-some-belonging-to-spacexsi/?ref=adjacent.media), some of it belonging to SpaceX. Coding agents need repository access, credentials, and context to be useful, so a security lapse exposes everything they touch. Any enterprise piloting agents should scope permissions per task, as the role-based access controls rolling out at larger firms do, and should ask vendors to show where context goes after a session ends.

## Commerce Rewired

**Publishers find two ways to charge for quality**

The Trade Desk's Zoë Mann argues that [publishers have new leverage as outcome-based buying beats cheap reach](https://www.beet.tv/2026/11/the-trade-desks-zoe-mann-publishers-find-new-leverage-as-outcomes-beat-cheap-reach.html?ref=adjacent.media). If advertisers pay for results, attentive audiences on quality sites earn more than bulk impressions do. Simon Owens examines the other lever: [whether news outlets should sell ad-free subscription tiers](https://open.substack.com/pub/simonowens/p/should-news-outlets-launch-ad-free). The two pull against each other. Ad-free tiers tend to remove the highest-value readers from the ad pool, which is exactly the audience outcome-based buyers want. Publishers losing search referrals need both revenue streams. They should price the ad-free tier with that tradeoff in view.

## Culture & Signal

**Medicare tests AI where denial is the product**

The EFF flags [Medicare's experiment with AI in coverage decisions](https://www.eff.org/deeplinks/2026/11/medicares-ai-experiment-effector-3820?ref=adjacent.media). The pilot brings algorithmic prior authorization into traditional Medicare, with vendors whose compensation reportedly tracks averted spending. A system rewarded for savings has a built-in reason to say no, and patients bear the cost of every false negative. Health-adjacent brands and insurers should expect audit and transparency demands to arrive before the efficiency gains are proven.

---

*17 articles across 6 fields · 15 sources · Powered by Folo + Claude*