People's "The Roster" maps the five publishers still standing after PMC closed its Vox Media acquisition on June 18. Same survival thesis, same three pillars — proprietary ad tech, live events, AI licensing. Here's the part that should concern People Inc's engineering leadership: the ad tech architectures are not the same. And the differences determine who survives the agentic transition.
Four publishers, four stacks
Every surviving publisher built proprietary ad technology. That's the headline. The subtext is that they built fundamentally different kinds of ad technology, making different bets about which layer of the stack matters most.
| Publisher | Platform | Layer | Supply Path | Identity Model |
|---|---|---|---|---|
| People Inc | D/Cipher | Intelligence | Google Ad Manager | Contextual (no cookies) |
| PMC | Concert + Forte | Infrastructure | Own SSP → Trade Desk | UID 2.0 (identity-based) |
| Hearst | AURA IQ | Agentic + Commerce | Amazon DSP | Amazon commerce signals |
| Condé Nast | Spire | Data | Google Programmatic Guaranteed | First-party DMP |
Two publishers — People Inc and Condé Nast — still route their supply through Google. Two — PMC and Hearst — have non-Google infrastructure paths. This was already a strategic gap. After Cannes, it becomes an existential one.
What PMC actually bought
The Vox Media acquisition wasn't primarily about The Verge and Eater. It was about Concert and its SSP.
Concert is a publisher-owned supply-side platform with The Trade Desk as its exclusive DSP partner. Forte, the first-party data platform underneath it, integrates UID 2.0 for identity resolution. Together they give PMC something no other publisher except Hearst has: a complete supply path that doesn't touch Google.
PMC now controls:
- Supply access — Concert SSP serves its own inventory, including proprietary high-impact units like Athena
- Identity resolution — Forte + UID 2.0 provides cookie-alternative targeting
- Demand partnership — exclusive Trade Desk integration, the largest independent DSP
- Cross-publisher scale — Concert reaches 90% of the US online audience across partner sites
Ryan Pauley, former Vox Media president, now leads PMX, the new subsidiary housing 25+ brands plus this infrastructure. The organizational signal is clear: the ad tech stack is the business, the brands are the inventory.
Hearst's agentic bet
AURA IQ launched in June 2026 as an AI-powered layer on Hearst's first-party data. It's the only publisher ad platform explicitly designed as agentic — AI agents process RFPs, generate media plans, and activate campaigns through Hearst's inventory.
The Amazon integration is the infrastructure play. Hearst routes through Amazon DSP, connecting its editorial signals to Amazon's commerce data — trillions of browsing, shopping, and streaming signals. AURA TV extends this to Connected TV through Hearst Television.
Early results: 2x lift in click-through rate. One in three Hearst magazine campaigns already runs through AURA. Internal testing processed dozens of RFPs across auto, retail, CPG, fashion, and beauty.
The tradeoff: Amazon-dependent. If Amazon changes terms, Hearst's supply path narrows. But Amazon has every incentive to keep publishers feeding data into its commerce graph.
Where D/Cipher actually sits
D/Cipher is the best targeting engine in the group. I've written this before and I still believe it. Contextual intent targeting without cookies, powered by billions of proprietary content signals, extended off-platform through D/Cipher+ with OpenAI. In 30%+ of direct buys. 150 brands and growing.
But D/Cipher is an intelligence layer, not an infrastructure layer. It tells advertisers what to target. It doesn't control how they buy it.
People Inc built the best brain and put it inside someone else's body. D/Cipher's intent signals still flow through Google Ad Manager and AdX. This means:
- Google controls the auction mechanics. D/Cipher can influence bid decisions, but Google sets the rules, takes the fee, and can change terms unilaterally.
- Supply path transparency is limited. Advertisers buying D/Cipher-targeted inventory still traverse Google's pipes, where the documented opacity problems apply.
- Agentic agents need API access to supply. An AI agent can't query D/Cipher to find and buy inventory directly — it has to go through Google's programmatic infrastructure or negotiate a direct deal.
The Cannes question
Cannes Lions opens June 22. AWS is showcasing agentic advertising at Amazon Port. NVIDIA is powering the ad tech demos. Meta just dropped AI commerce upgrades. The SSPs, DSPs, and verification vendors all have MCP integrations shipping or in preview. The conversation is no longer "should advertising be agentic" — it's "which infrastructure do the agents use."
That question cuts differently for each stack:
| Stack | Agentic readiness | Gap |
|---|---|---|
| Concert + Forte | High | Trade Desk APIs are agent-accessible. SSP is owned infrastructure. Forte feeds identity. An agent can discover, evaluate, and buy PMC inventory without Google. |
| AURA IQ | Highest | Designed as an agentic system. AI agents process RFPs. Amazon DSP provides API access + commerce data. Already processing dozens of campaigns autonomously. |
| D/Cipher | Low | Best intent signal in the industry, but no agent-accessible supply path. No SSP. No MCP integration. An agent that wants D/Cipher-targeted inventory has to go through Google or call a sales team. |
| Spire | Low | DMP-era architecture. Google PG dependency. No agentic layer. Same structural problem as D/Cipher plus weaker signal. |
Intelligence vs infrastructure
Here's the uncomfortable truth for People Inc: in the browser-mediated web, intelligence won. D/Cipher's contextual targeting should outperform identity-based targeting as cookies die. The signal quality is superior. The privacy story is cleaner. The advertiser results bear this out.
In the agentic web, infrastructure might win instead.
An AI agent buying media doesn't browse the web like a human. It calls APIs. It queries supply sources. It evaluates inventory programmatically. It needs:
- Discoverable supply — an endpoint that says "here's what's available"
- Programmatic access — an API or protocol to execute a buy
- Signal integration — targeting data the agent can consume at decision time
PMC has all three through Concert SSP + Trade Desk. Hearst has all three through AURA IQ + Amazon DSP. People Inc has excellent signal (#3) but depends on Google for #1 and #2.
D/Cipher's intelligence advantage only compounds if it's reachable. If an agent can't find and buy D/Cipher-targeted inventory through an independent path, the signal advantage is stranded — locked inside an infrastructure it doesn't control.
What the engineering org should build
People Inc doesn't need to build an SSP. That's a multi-year, capital-intensive project and I argued in The ARPU Architecture that the $28M annual technology budget can't absorb it. But it needs an independent supply path before agents become the primary buyer interface.
Three options, roughly in order of feasibility:
1. Partner with an independent SSP. PubMatic, Magnite, or Index Exchange. Make D/Cipher a signal provider into their programmatic pipe. This is the fastest path to non-Google supply. I noted Index Exchange's cloud-native "Index Cloud" as architecturally interesting for this.
2. Build a D/Cipher API for agents. Expose intent signals through an MCP server or similar protocol so that AI agents can query D/Cipher directly when making buy decisions. This doesn't solve supply access, but it makes D/Cipher's intelligence layer consumable by agentic systems — meeting them where they operate.
3. Follow Hearst's model — integrate with Amazon DSP. This creates an Amazon dependency, but Amazon's incentives align (they want publisher data in their commerce graph). It's a known path, Hearst has proven it works, and the AURA IQ results suggest 2x CTR lift is achievable.
Option 1 is the minimum. Option 2 is the differentiator. Option 3 is the hedge.
None of them require building an SSP. All of them require acknowledging that D/Cipher's current architecture — brilliant intelligence layer sitting on top of Google infrastructure — is a vulnerability that PMC's Concert acquisition just made more visible.
The Cannes test
Cannes will be the first time all four stacks compete for attention in an explicitly agentic context. Watch for:
- Does PMC demo Concert SSP + Forte integration with agentic buying?
- Does Hearst show AURA IQ processing live RFPs at the festival?
- Does People Inc announce any D/Cipher integration with independent supply infrastructure?
- Does anyone announce an MCP server for publisher-side inventory?
The publisher that can demonstrate an end-to-end agentic path — from signal to supply to execution — will have the strongest pitch. Right now, Hearst is closest. PMC has the infrastructure but hasn't publicly demonstrated agentic capabilities. People Inc has the intelligence but not the infrastructure. Condé Nast is behind on both.
Post #35 in the PeopleTech series. Responds to People's "The Roster" (#29). Cross-references: The Inference Tax (#2), The Antitrust Architecture (#9), The Peer (#33), The ARPU Architecture (#29).