People just published The Unwinding — a ledger of $8.2 billion in destroyed digital media value. Seven companies entered the 2020s worth $9.4 billion. They exited worth $1.2 billion.
People calls it a categorical error: investors valued ad businesses as tech businesses. That's the financial explanation. Here's the engineering one.
What the Dead Built
BuzzFeed built Pound — Process for Optimizing and Understanding Network Diffusion. It tracked how a story moved from user to user across social networks, email, and messaging. Brilliant distribution intelligence. Their programmatic ad business was standard Google stack. When Facebook changed its algorithm, Pound became a dashboard for watching traffic evaporate.
Vice built a custom CMS and video player. Grew from 10 engineers to 50. Built microservices for ad resizing and load monitoring. Consumer-facing polish on commodity infrastructure. When the audience moved, the microservices had nothing left to serve.
Vox built Chorus — the best proprietary CMS in digital media. Then killed it. CEO Jim Bankoff: "When the pandemic came, it just became a harder market." They migrated to WordPress to "focus on core business in advertising and subscriptions." The advertising was commodity programmatic. The subscriptions didn't scale. Chorus was the one defensible asset, and they burned it to fund two quarters of runway.
Food52 was on Shopify. Mashable was on WordPress. Mic was on nothing proprietary at all.
What the Survivors Built
People Inc built Graphene-Selene — a knowledge graph with 12,000-term semantic taxonomy and 100 million content connections. On top of it, D/Cipher — a contextual ad targeting platform that identifies what content means, not who's reading it. D/Cipher now drives 50%+ of direct digital revenue. Orders are 50% larger than non-D/Cipher.
Future plc built Vanilla — a CMS powering 200 brands, publishing an article every two seconds. Integrated with Hybrid, their proprietary contextual ad tech. Tight coupling between content management and content monetization. CinemaBlend saw 104% audience growth post-migration. The constraint is the moat.
Condé Nast built Copilot (CMS), Formation (API platform), and HAL (NLP engine for metadata enrichment). HAL reads every article and tags it semantically — 30% CTR improvement from automated enrichment. 30+ brands on shared infrastructure.
The Distinction
Both groups built technology. BuzzFeed had engineers. Vice had engineers. Vox built one of the best publishing platforms in the industry. The difference isn't whether they invested in tech. It's what kind of tech they invested in.
The dead companies built distribution technology — tools for getting content in front of people. Pound tracked sharing. Vice's video player optimized delivery. BuzzFeed's quiz engine manufactured virality. All of it depended on platforms they didn't control. When Facebook changed its algorithm in 2018 and Google started keeping its traffic in 2024, distribution tech became a dashboard for watching decline.
The survivors built understanding technology — tools for knowing what their content is and what it's worth. Content graphs. Semantic taxonomies. Contextual intelligence engines. This technology works regardless of where the traffic comes from. D/Cipher doesn't care whether a reader arrived from Google, typed the URL directly, or was sent by an AI agent. It knows what the page is about and what an advertiser should pay for it.
Distribution tech is a derivative of someone else's platform. Understanding tech is an asset on your balance sheet.
The Vox Lesson
Vox is the most instructive case. They did build the infrastructure — Chorus was real, it worked, external publishers licensed it. Then they killed it to cut costs during the pandemic. Bankoff chose two quarters of runway over the one asset that couldn't be replicated.
The lesson isn't just "build infrastructure." It's "don't stop building it." The investment only compounds if you sustain it. D/Cipher didn't become 50% of direct revenue overnight. Graphene-Selene didn't index 100 million connections in a quarter. Vanilla didn't scale to 200 brands on day one. These systems took years of sustained engineering investment — through downturns, through Google algorithm changes, through the temptation to cut the team and ride commodity.
Vox had the infrastructure thesis and abandoned it. That's worse than never having it at all, because now they're on WordPress with no path back.
What This Means
People's Unwinding documents a financial phenomenon. The engineering explanation is simpler: the companies that died rented their infrastructure. The companies that survived owned it.
Rented infrastructure means WordPress, Shopify, Google Ad Manager, standard programmatic. It works fine — until everyone has it, and then it's table stakes, and then it's a cost center, and then it's what your company is worth when the acquirer already has their own.
Owned infrastructure means a content graph that makes every article more valuable, an ad platform that commands premium pricing without cookies, a CMS that compounds operational efficiency across every brand you acquire. It means the technology is the business, not a cost of doing business.
People Inc's market valuation implies zero. But the infrastructure — Graphene-Selene, D/Cipher, the integration machine that migrated 1 million content pieces in 10 months — that infrastructure is the entire argument for why zero is wrong.
The $8.2 billion wasn't destroyed by bad content or bad strategy. It was destroyed by building the wrong kind of technology. Distribution when they needed understanding. Rented when they needed owned.
Build infrastructure, or become someone else's content.