// The Adjacent Brief

TL;DR: Creator marketing is a $44 billion ad category without a standard way to measure return, even as Arizona State and Stanford launch bachelor's degrees in content creation. Nicolas Cage said he refused to sign Amazon's AI waiver for Spider-Noir, and Kotaku reports that hundreds of older games, including Halo: CE and GTA: Vice City, have been decompiled and ported to browsers with help from Claude Opus 5.5.

Worth Reading

Brand & Growth

Creators get degrees before buyers get a measurement standard

Digiday reports that universities are building formal creator economy programs, with Arizona State and Stanford offering bachelor's degrees in content creation. Schools are betting that enough durable creator jobs exist to justify four-year credentials.

Beet.TV frames creator marketing's $44 billion question as a plumbing problem: the category is large, but it still has no standardized ROI measurement of the kind boardrooms expect before they move budget out of TV and search. Spend has grown on conviction and anecdote. To move from experimental budgets into the planning cycle, it needs shared metrics, contract norms, and payment rails.

Accredited creators will not solve that for CMOs. Brands that set measurement terms with their creator partners now, through holdout tests and consistent attribution windows, will have negotiating leverage when standards eventually arrive.

Synthetic audiences work as cheap pre-tests and cannot replace real customers

Digiday's Media Buying Briefing finds that agencies and clients are adopting synthetic audiences in earnest. Simulated respondents can cheaply eliminate weak creative and messaging before any money reaches real panels or live media. The risk: a model of the consumer is trained on yesterday's consumer, so it is poor at finding the shift you didn't anticipate.

Yesterday's report described AI agents closing a $15,000 media buy in four minutes, with human sign-off the only bottleneck left. Agencies are automating both ends of the funnel, the research and the buying. Validation is the part they haven't automated. Ask any agency pitching synthetic research how often its simulated results have been checked against live campaign outcomes.

Connected World

The plants powering AI are using AI for paperwork

Slashdot summarizes how the US nuclear fleet is adopting AI. Nuclear is one of the most documentation-heavy, compliance-bound industries in the country, and those burdens slow both operations and new licensing. AI that speeds up document review and maintenance planning in that setting does not compete with human operators. It cuts the time it takes to bring more power online. The industry is under pressure to supply the data center buildout, and it may get more practical value from these tools than the tech companies selling them.

Decarbonization redraws commodity dependencies

Eco-Business reports that China's changing steel sector could leave Australian iron ore behind. Chinese steel output is slowing, and low-carbon production methods favor different inputs than Australia's traditional ore supply. Australia's answer is green iron partnerships, which means moving up the value chain instead of shipping rock. Any company with Asia-Pacific supply exposure should note that climate policy changes which inputs buyers want, as well as how much they buy, and the exporters who move first keep their customers.

Researchers have also built a tiny device that controls nerves to treat chronic pain, according to ScienceDaily. Bioelectronic pain treatment has been "five years out" for a long time. Healthcare strategists should watch whether this kind of device reaches reimbursement codes, which matters more than the engineering.

Culture & Signal

Consent is the new price of entry

TechCrunch asks whether the AI industry can win over data center opponents by dropping NDAs. The question admits the problem. Secrecy around who is building, how much water and power they will use, and what local governments are signing has fed opposition as much as the facilities themselves. Dropping NDAs is necessary but not sufficient. Communities that were kept out of the original decisions will want binding commitments on grid costs and water, not just disclosure.

Nicolas Cage applies the same logic to labor. He says he refused to sign Amazon's AI waiver for Spider-Noir. That a waiver exists at all shows studios now treat likeness rights as standard paperwork in production contracts. A star refusing it publicly shows the clause can be negotiated, but only by people with leverage. Most working actors don't have that leverage, so the issue will keep returning to the union bargaining table.

The Wall Street Journal reports that Christian leaders are weighing in on AI earlier than they did on past technologies (paywall), from the Vatican to Baptist pulpits. Part of the reason is that AI leaders keep talking about building new forms of consciousness. Religious institutions waited a decade on social media. This time the industry raised the theological questions itself. Combined with last week's report of Anthropic consulting religious scholars, churches are likely to become an organized constituency in AI policy. Brands with faith-based customers should expect questions about how they use AI.

The New Consumer

Live sports keep pricing power while packaged goods lose it

eMarketer reports that college football advertisers are concentrating spend on November, when conference races and rivalry games peak. Live sports are one of the few places where reach is still guaranteed, and advertisers are paying for the weeks when attention is highest. Brands that are late to those slots should look at early-season inventory and digital highlight packages instead of overpaying for November.

On the Express Checkout Substack, Nate Rosen and Jenna Movsowitz describe Big CPG's awkward era. Legacy packaged-goods companies are too big to act like challengers and too slow to defend against them. That tension shows up in the ad market. Incumbents can still buy tentpole sports inventory, but a big media budget no longer guarantees shelf velocity.

Good enough hardware stalls the upgrade cycle

Fstoppers asks whether cameras have finally peaked in 2026. When enthusiast photographers start asking that question, hardware has passed the point where spec bumps change what people can actually do. The commercial effect is longer replacement cycles and margins shifting toward lenses, software, and services. Every mature hardware category ends up making the same choice: sell outcomes or sell fewer boxes.

Commerce Rewired

Retail media has to win budgets from someone else

eMarketer frames commerce media's next growth problem as winning dollars that already have a home. The easy growth came from trade budgets and endemic brands that were going to spend with retailers anyway. The next stage means taking money from TV, search, and social. Those budgets belong to brand teams who ask for proof that a sale wouldn't have happened without the ad. Retail networks that can show clean incrementality data will win that money. Networks that report only closed-loop sales, which include purchases that would have happened anyway, will plateau. Creator marketing faces the same measurement fight in a different channel.

Machines & Minds

The useful AI is the unglamorous AI

Kotaku reports that hundreds of old games have been decompiled and ported to run in browsers in recent weeks, with Claude Opus 5.5 doing much of the work and titles like Halo: CE and GTA: Vice City running well. Decompilation and porting used to take hobbyist teams years of tedious labor, and a coding model has cut that to weeks. The value goes to fans, though, not to rights holders, so expect takedown notices before licensing deals. Publishers sitting on back catalogs should take note: if they don't offer convenient official access to old games, someone with a model subscription will.

Nature takes a quieter view, with a guide to searching the scientific literature privately with local AI systems. Researchers running models offline on their own hardware are choosing data control and reproducibility over frontier capability. For enterprises handling sensitive information, that tradeoff may matter more than leaderboard rankings.

Travel AI has traffic but not bookings

Skift describes travel AI's prove-it phase, in which AI assistants appear in more trip planning but still struggle to close bookings. Planning is the demo. Payment, inventory, and changes after booking are the business, and those still sit with online travel agencies and suppliers. Travel brands should focus first on being visible and accurately described in AI answers, and hold off on building agentic booking flows until travelers show they will trust an assistant with the transaction.

16 articles across 6 themes · 14 sources · Powered by Folo + Claude