A bungalow is not a badly built house. The foundation is solid. The walls are plumb. The craftsmanship is often excellent. A bungalow is a design decision: build one story, build it well, and stop.
People Inc is building a bungalow.
The Floor
People Inc's defensible revenue is growing. Affiliate commerce rose 26% year over year in Q1 2026 — driven by category mix in home, food, and lifestyle where AI Overviews haven't disrupted the way they've disrupted tech product reviews. Food & Wine Classic Aspen just completed its 43rd year: 5,000 attendees, $2,950 tickets, roughly $15 million from a single weekend with zero Google dependency. The company acquired Hot Luck Austin in June, plans 60+ events for 2026, and has AmEx/Resy sponsorship locked through 2028.
This is real revenue from real relationships. Commerce Producers insert affiliate links by hand, selecting products they've actually tested. Event teams book chefs, negotiate venues, handle health code compliance per city. Each new event requires incremental operational headcount. Each new affiliate recommendation requires an editor choosing, testing, and writing about a specific product. There is no network effect. The 469:1 ARPU ratio between event attendees and digital sessions is real, but it only applies to humans physically present.
The floor is bounded. Not small — the food vertical alone is probably $100M+ combined — but bounded by human curation, physical venues, and editorial labor. It grows by addition, not multiplication.
The Ceiling
Above the floor sits D/Cipher — the strongest contextual intelligence platform in consumer publishing. 100 million+ semantic connections, OpenAI-enriched taxonomy, pre-computed embeddings across ~200 million monthly uniques and 40 brands. Orders including D/Cipher are 50% larger. It drives over half of direct digital revenue. Jim Lawson was hired to expand it to third-party sites and CTV.
Beside it sits Graphene-Selene — the proprietary content graph powering all 40 brands. GraphQL APIs, Spring microservices, graph database architecture. The most comprehensive knowledge graph in consumer publishing, connecting every recipe, product review, health article, and travel guide into queryable entities.
Both systems are the best in their category. Both are completely closed.
D/Cipher has zero AAMP participation, zero ARTF containerization, zero MCP exposure, zero Agent Registry entries. Graphene-Selene has zero MCP servers, zero agent discovery protocols, no way for an external system to query its content graph. The intelligence is genuine. The interface is absent.
One Pattern, Two Systems
The standard reading is two technology gaps that need two fixes: build an MCP server for D/Cipher, build one for Graphene-Selene, ship them, move on. That reading misses what makes this structural.
These aren't two independent technology decisions. They're the same organizational pattern expressing itself in two domains.
D/Cipher operates through bilateral deals — direct insertion orders where advertisers get targeting data through a managed service. Graphene-Selene operates through bilateral partnerships — the OpenAI content licensing deal, the Microsoft à la carte arrangement, the Meta multi-year agreement. Both systems take their strongest asset and expose it only through controlled, one-to-one relationships.
This is the same instinct that makes the floor excellent. Food & Wine Classic works because someone curates the 40 wines, selects the presenting chefs, manages the AmEx relationship personally. Commerce recommendations grow because an editor chooses each product. Quality requires control. Control requires bilateral relationships. Bilateral relationships don't compound.
Meanwhile, the industry moved to protocols. On the CMS side: Sanity shipped 40+ MCP tools for agent-driven content management. Brightspot built governance-scoped self-discovery for agents. WordPress 7.0 made MCP a flagship feature. Storyblok, Hygraph, Strapi, Payload, Directus, Contentful, Kontent.ai — eleven platforms and counting, each enabling any agent to query, create, and manage content. Graphene-Selene: zero.
On the ad tech side: AAMP 3.0 ships in July with a working cloud reference implementation. The ARTF spec has MCP built directly into it. PubMatic Decision Fabric runs GPU-accelerated agent models inside the auction at ~1ms. Index Cloud offers containerized compute for buyer intelligence. Magnite Orchestration coordinates buyer-seller agents across premium inventory. The Kochava StationOne AAMP workspace is in open beta. D/Cipher: zero entries, zero participation.
Same organizational instinct. Same result. The taste for curation that makes a $15 million food festival excellent is the same taste that keeps a $500 million ad platform from shipping an MCP server.
Platform or Lobby
The Inversion strategy — turning brands into products — was supposed to build the second story. D/Cipher was the platform layer connecting 40 brands into a unified intelligence surface. If D/Cipher opens, it compounds: 200 million uniques across home, food, health, travel, entertainment, beauty, queryable by any buyer agent. That's a contextual graph no single-vertical publisher can match.
If D/Cipher stays closed, Inversion is the company. Nine brands becoming products — People app, Food & Wine events, Allrecipes identity layer, InStyle comeback — each growing independently, each requiring independent operational investment. Each is its own bungalow.
There is a third option, and it's been sitting on the MGM board's desk for 28 days at $48.30 per share.
If the floor can't scale and the ceiling can't hold, the MGM bid is an architectural decision, not just a financial one. Buy a skyscraper. Put the publishing operation in the lobby. MGM's casino EBITDA — roughly $800 million annual share from Las Vegas, Macau, and eventually Osaka's $10 billion integrated resort — becomes the ceiling that D/Cipher couldn't build. The publishing floor keeps doing what it does well: curating content, running events, powering affiliate commerce. Just not as the main business.
Whether that's strategic genius or strategic surrender depends on timing. MCP adoption is on a 12-18 month arc to becoming assumed standard — the same trajectory Prebid followed from 2015 to 2018. The MGM deal, if approved, closes sometime in 2027. If D/Cipher re-engages with agentic infrastructure after the deal closes, the protocols may have already calcified without it. If it doesn't re-engage, the question becomes moot. People Inc becomes a casino holding company with a content division.
The Craft
A bungalow is not a failure of construction. People Inc's floor is the strongest in publishing — commerce growing, events expanding, editorial quality that consumers trust 300% more than AI-generated alternatives. The craftsmanship is excellent.
But eleven CMS platforms built second stories this year. Three SSPs built entire skyscrapers of agentic infrastructure. The IAB standards body shipped the blueprints. People Inc's two most sophisticated technology systems — the ones that were supposed to be the ceiling — remain bilateral, curated, closed. Not because the engineers can't build the protocol layer. Because the organization builds everything the way it builds events: one relationship at a time, quality-controlled, hand-selected.
That works for a floor. It doesn't work for a ceiling.