> ## 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 — October 5, 2026
- URL: https://adjacent.media/briefs/2026-10-05/
- Published: 2026-10-05T14:15:03.000Z
- Updated: 2026-10-05T14:15:03.000Z
- Description: Meta is building one of the world’s largest undersea cable networks, adding to seafloor holdings that span six continents.
- Author: Jonathan Greene
- Tags: #brief

**TL;DR:** Meta is building one of the world's largest undersea cable networks, adding to seafloor holdings that span six continents. A federal judge called Flock's license-plate reader system "indiscriminate mass surveillance," and The New York Times reports that brands are rewriting their sales pitches for AI agents that shop on consumers' behalf.

## Worth Reading

- [Shipping routes around the Gulf war through a melting Arctic](https://www.ft.com/content/2e0eb698-d4d3-4bcc-927a-2997df7709be?syn-25a6b1a6=1&ref=adjacent.media) (paywall)
- [AI seat cuts show up at renewal before they show up in your forecast](https://open.substack.com/pub/thecscafe/p/ai-seat-reduction-at-renewal)
- [A federal court treats Flock's scale as the constitutional problem](https://techcrunch.com/2026/10/03/federal-judge-calls-flock-indiscriminate-mass-surveillance/?ref=adjacent.media)
- [Agents let you stop paying for the parts of software only humans used](https://open.substack.com/pub/natesnewsletter/p/ai-agents-software-buying)
- [Robot caregivers may be a question of preference, not just staffing](https://marginalrevolution.com/marginalrevolution/2026/10/do-the-elderly-prefer-robotic-care.html?utm%5Fsource=rss&utm%5Fmedium=rss&utm%5Fcampaign=do-the-elderly-prefer-robotic-care)
- [llms.txt asks site owners to build a second front door for machines](https://lagomor.ph/2025/08/llms.txt/?ref=adjacent.media)
- [Meta is buying its way out of renting the internet's backbone](https://fortune.com/2026/10/03/meta-building-undersea-cables-petal-us-france-mark-zuckerberg/?ref=adjacent.media)

## Connected World

**Meta wants to own the pipes**

Fortune reports that Meta is building [one of the world's biggest undersea cable networks](https://fortune.com/2026/10/03/meta-building-undersea-cables-petal-us-france-mark-zuckerberg/?ref=adjacent.media), part of a seafloor portfolio that reaches six continents. For most of the internet's history, subsea cable was a consortium business. Carriers and governments split the cost, and tech companies bought capacity. Meta building at this scale changes who controls capacity pricing, routing and repair priority on the routes that carry its traffic.

The open question is what the spend buys. Meta is not a bandwidth reseller, so the return comes from lower long-run transit costs and more predictable latency as AI workloads move between data centers. It also gains leverage in negotiations with the carriers it no longer depends on. Telecoms that relied on hyperscalers as anchor customers on new routes now have a customer turning competitor on the seafloor. Governments should expect more cable-landing decisions to be negotiated with a single company rather than a consortium.

**Exoskeletons earn their keep at work before they earn it on trails**

Ars Technica argues we are seeing [the dawn of the age of the exoskeleton](https://arstechnica.com/science/2026/10/the-dawn-of-the-age-of-the-exoskeleton/?ref=adjacent.media), with powered assist suits moving beyond rehab clinics and toward everyday use. The utility test matters here.

\- **Where a value loop exists:** A suit that reduces lifting injuries on a warehouse floor or keeps an aging worker on the job longer pays for itself in workers' compensation claims and retention.

\- **Where it does not yet:** A suit sold to weekend hikers is a gadget until battery life, heat and weight stop being the first things owners mention.

Brands looking at the category should watch employer purchase orders and insurer interest, since consumer launch sizzle reels reveal far less about where the category is headed.

## The New Consumer

**Gen Z is paying for certainty by the reading**

Fortune reports that [Gen Z is paying witches and tarot readers](https://fortune.com/2026/10/04/gen-z-witches-tarot-readers-psychic-services/?ref=adjacent.media) to make an uncontrollable future feel manageable, and Etsy has become one of the marketplaces carrying that demand. The evidence is behavioral: people are spending money, often repeatedly. The spending looks like a working subscription model for reassurance, though it is easy to dismiss as quirky. It is cheap, personalized and ritualized, and it fills the gap left by institutions young consumers don't trust to deliver a stable career or housing path.

For brand strategists, the takeaway is not "add astrology." Consumers will pay a premium for a sense of agency, and categories built around prediction and control are where that money is landing. That includes financial planning apps, career tools and even wellness.

**Your friends didn't sign up for Muse**

WIRED profiles [Muse, which builds detailed profiles of your friends and family](https://www.wired.com/story/muse-creates-detailed-profiles-of-all-your-friends-and-family/?ref=adjacent.media) so a personal agent can manage your relationships. The user opts in, but the people being profiled do not. Every privacy debate in consumer tech so far has been about what a company knows about its own users. Muse turns the user into the collection mechanism for everyone around them. Expect regulators and plaintiffs to look there first.

Michael Chadwick's essay on [the binge-purge tech cycle](https://michaelchadwick.info/blog/2026/01/23/the-binge-purge-tech-cycle/?ref=adjacent.media) explains why products like Muse get adopted anyway. Users pile on tools, feel buried by the complexity, then cut back hard, and companies design for the binge phase. Product leads should note that retention is decided in the purge. Tools that don't justify themselves within a few weeks get cut in the clean-out.

## Culture & Signal

**The ruling targets Flock's design**

A federal judge described Flock's license-plate reader network as ["indiscriminate mass surveillance"](https://techcrunch.com/2026/10/03/federal-judge-calls-flock-indiscriminate-mass-surveillance/?ref=adjacent.media) and found that warrantless searches of the system violate the Fourth Amendment, TechCrunch reports. The language matters. Most surveillance challenges argue that a tool was abused. This one goes after the product's core feature: blanket capture that police can query without a warrant.

Flock's growth depended on a regulatory gap. Police departments could get searchable location histories without going through a judge, and private customers such as homeowners' associations and retailers expanded the camera network at no cost to law enforcement. If the reasoning holds on appeal, every organization feeding that network inherits some of the liability. Legal and comms teams at companies with cameras in their parking lots should be asking where that footage goes.

**A rewrite is not a relicense**

Eddy Luten documents how a copyleft project can be license-laundered, using the chardet character-encoding library. The library moved from a copyleft license to a permissive one through a rewrite of the codebase, raising the question of whether a rewrite, especially an AI-assisted one, clears the original obligations. If it does, copyleft stops protecting anything: anyone with a capable model can rewrite a project and relicense it.

Companies shipping software should treat dependencies with this kind of history as having unresolved provenance until courts or foundations say otherwise.

## Brand & Growth

**Brands are now writing for a buyer with no impulses**

The New York Times reports that [as AI agents begin shopping, brands are changing their sales pitch](https://www.nytimes.com/2026/10/03/business/dealbook/ai-agent-marketing.html?ref=adjacent.media) (paywall). Content is being optimized for software that compares specs, prices and return policies, not for a person who responds to mood and aspiration. An agent doesn't care about your hero video. It cares whether your product data is structured, complete and consistent across retailers.

Brand building remains relevant, but it moves earlier in the purchase. Emotional brand work earns the moment when a person tells the agent "get me the Patagonia one." Product-page work decides whether the agent can find, verify and recommend the product when the person doesn't name a brand. A CMO should run both audits:

\- **For the human:** Is the brand strong enough that people name it in the prompt?

\- **For the machine:** Can a shopping agent read and compare the catalog without guessing?

## Commerce Rewired

**Agents unbundle what SaaS sold as one price**

Nate's Substack argues that [agents split your software into three parts](https://open.substack.com/pub/natesnewsletter/p/ai-agents-software-buying) that deserve different spending decisions. The core of the argument is that the layer an agent needs (the data and the business logic) has different value from the interface built for humans to click through. Per-seat pricing bundled them together.

Vendors already see the problem. Several enterprise software makers are adding separately priced tiers for agent features, which fragments what used to be one subscription. Buyers walking into renewals should ask which layer they're paying for, and whether the interface fee still makes sense once fewer humans log in.

## Machines & Minds

**The pitch of one agent to run your life has yet to become a product**

Grok, Muse and OpenAI have all launched personal agent products within roughly two months. Gettheleverage pushes back on the premise behind all of them in [an argument against the personal-agents theory of everything](https://www.gettheleverage.com/p/against-the-personal-agents-theory?ref=adjacent.media). A single agent mediating your email, shopping, calendar and relationships is an elegant vision, but most people don't have a single bottleneck that one agent fixes. Each delegated domain also adds a new way for the agent to fail on your behalf.

The utility test applies. Agents that own one well-scoped, repeatable job, like expense filing, rebooking travel or procurement comparisons, have a value loop you can measure. The do-everything assistant mostly works as a fundraising story and a way to collect user data, which is how Muse's relationship profiling ends up in a product pitch. Strategists deciding where to build or partner should bet on the narrow agents.

---

*11 articles across 6 themes · 10 sources · Powered by Folo + Claude*