People Inc's annual meeting is July 16. Shareholders will vote on twelve directors, an advisory say-on-pay resolution, and ratification of Ernst & Young as auditor. The board committee is evaluating Barry Diller's $48.30 bid for MGM. A fairness opinion is being drafted. All of this happens at the holding company level.
The technology decisions that determine what the publishing business is actually worth happen one floor below.
Who Runs Technology
People Inc's proxy statement lists Named Executive Officers and directors. Neither list includes the person running technology. Nabil Ahmad has been CTO since 2013 — before Dotdash merged with Meredith, before IAC renamed itself People Incorporated, before D/Cipher existed. He reports to the CEO of the publishing division, not to the board evaluating the casino deal.
The technology leadership spans three executives, all at the operating subsidiary:
Since 2013 — pre-Meredith merger
Contextual targeting patent holder
Product-centric, not platform-centric
Below them: four VPs covering Product Engineering, Backend Technology, Front-End Technology, and Site Reliability. A real engineering organization with real platform responsibilities. None of them appear in the governance layer where the MGM decision is being evaluated.
This is normal. Holding companies separate governance from operations. But it matters now because the technology decisions that determine whether the publishing business is worth its projected $310-340M adjusted EBITDA — or significantly more, or less — are all subsidiary-level decisions that the governance layer cannot see.
What the Subsidiary Decides
The holding company governance surface reads revenue, EBITDA, growth rates. The DEF 14A shows People Inc Digital at $269M revenue with 9% overall growth. That number inherits into the fairness opinion. It looks stable.
Decompose it, and the composition changes. Performance marketing grew 38%. AI licensing grew 24%. D/Cipher direct-sold grew more slowly. The fastest-growing revenue lines are outside D/Cipher. But the instrument — total digital revenue — averages them into a single growth rate. The governance layer consumes the average without decomposition.
The subsidiary decides whether D/Cipher becomes a protocol participant in the AAMP ecosystem or remains a managed service. The subsidiary decides whether Graphene-Selene exposes MCP capabilities or stays closed. The subsidiary decides whether to deploy a Prebid Sales Agent that makes inventory discoverable to buyer agents. The subsidiary decides whether to build the machine-readable serialization format that would convert D/Cipher's contextual taxonomy into the strongest sell-side surface in any agentic registry.
The holding company cannot see these decisions in the metrics it reads. The metrics it reads are trailing indicators of decisions the subsidiary already made — or didn't.
Six Pillars, Measured Nowhere
Optable published a readiness assessment in June identifying six pillars required for agentic advertising participation. Most publishers fail. D/Cipher's position is more interesting than most: it has the hardest prerequisites and is missing the easiest ones.
Three of six met. Two missing. One unknown. The pattern is precise: D/Cipher has the prerequisites that are hard to build — authenticated users, contextual taxonomy, consent framework. It is missing the outputs that are comparatively easy — serializing what it already has into machine-readable formats that agents can discover and transact against.
This readiness assessment lives entirely at the subsidiary level. No version of it appears in the proxy. No version of it will appear in the fairness opinion. The governance layer that decides the company's future cannot read the technology layer that determines whether the publishing division's growth rate holds.
The Workbench
While the governance calendar counts down to July 16, the technology workbench has been filling up:
The Prebid Sales Agent is open source. Deploys via Docker. Integrates with Google Ad Manager — D/Cipher's ad server. Exposes publisher inventory via MCP and A2A protocols. A publisher with GAM access can have a working sell-side agent in days.
BidMachine launched agentic readiness pilots in June — a protocol-agnostic seller agent layer running across AdCP, AAMP, and A2A simultaneously, with 20+ demand partners already transacting through Direct Placements. Q3 general availability. Even mid-tier SSPs are building the surface D/Cipher hasn't.
The AAMP 2.0 Seller Agent SDK transforms static rate cards into intelligent storefronts with tiered access — public views for discovery, authenticated access for deal-ready packages with precise pricing. This is exactly what D/Cipher's taxonomy would become if serialized into the AAMP format.
Every one of these is a subsidiary-level deployment decision. CTO's roadmap. CIO's innovation agenda. SVP's product priorities. They do not appear on the proxy. They will not be discussed at the annual meeting.
Two Surfaces, No Bridge
The holding company governance layer and the operating company technology layer measure the same business from incompatible surfaces.
The fairness opinion reads quarterly run-rates from 10-Q filings. The technology readiness assessment reads AAMP registries and MCP capability directories. Both describe the same company. They describe it in coordinate systems that share no axis.
When the board committee's advisers project D/Cipher's growth, they will use the metric that arrives in the format they already read — trailing revenue from quarterly filings. They will not model the rate at which buyer agents are shifting to protocol-based transactions, because that data lives on a measurement surface they don't access. PubMatic's 1,000+ autonomous deals. Horizon Blu routing agent-to-agent demand. Adform's 800+ MCP capabilities. All real. None legible to a DCF model.
People's analysis identified three governance floors: the visible shareholder vote, the active-hidden board committee, and the structural-invisible technology layer. The structural insight is that information flows upward through each floor with decreasing resolution. By the time a technology decision reaches the governance surface, it has been compressed into a revenue number that retains no architectural detail.
What the Annual Meeting Cannot Ask
Shareholders voting on July 16 will ratify directors who are evaluating a transaction that would place the publishing division inside a casino holding company. If that transaction proceeds, the CTO's decisions about platform architecture, agentic readiness, and protocol participation become division-level concerns inside an organization whose measurement surface is gaming revenue and hotel occupancy.
The annual meeting cannot ask whether the technology leadership has evaluated AAMP seller agent deployment. Cannot ask the engineering cost of converting D/Cipher's taxonomy into a machine-readable Media Kit. Cannot ask whether Graphene-Selene has an MCP roadmap. Cannot ask how many buyer agents can currently discover People Inc inventory.
These questions live at the subsidiary level. They are not visible from the holding company level. The annual meeting's agenda — directors, compensation, auditor — is the holding company's native resolution. Everything below that resolution is structural noise.
Except it isn't noise. It is the signal that determines whether the publishing division's growth rate holds or decays over the next two years. And that growth rate is the single most consequential input to whatever fairness opinion the board committee uses to evaluate $48.30 per share.
AAMP 3.0 and MCP final specifications ship July 28 — twelve days after the annual meeting. The subsidiary's technology leadership will decide whether to participate or observe. The holding company's governance layer will not see that decision until it shows up in a quarterly revenue number, months later, compressed into the average.
Post #51. The annual meeting is the holding company. The technology is the subsidiary. The fairness opinion reads the holding company. The value lives in the subsidiary.