The Biggest Bundle?

Key Takeaways

• Newspapers succeeded by bundling diverse content.
• Internet unbundled news, causing massive newspaper decline.
• AI currently follows bundling model, but may be unbundled.
• Job impact hinges on weak vs strong bundle occupations.
• Apple uses model distillation for on‑device AI specialists.

Summary

The article traces how newspapers became dominant by bundling diverse content types—news, sports, lifestyle, and ads—leveraging continuous‑feed presses, the telegraph, and rail distribution. The internet later unbundled each component, precipitating a 45% drop in U.S. newspaper circulation and the loss of thousands of jobs. It argues that artificial intelligence now exhibits similar bundling traits, but model‑distillation efforts, such as Apple’s on‑device versions of Google’s Gemini, could unbundle AI into task‑specific, privacy‑focused tools. Ultimately, the impact of AI on employment depends on whether occupations are weakly or strongly bundled.

Pulse Analysis

The newspaper’s rise illustrates classic bundling economics: a single product combined news, entertainment, and advertising, turning the press into a profit engine. Continuous‑feed rotary presses slashed marginal printing costs, while the telegraph supplied instant content and railroads ensured nationwide distribution. By offering a one‑stop information bundle, publishers maximized circulation and ad revenue, cementing the medium as a societal cornerstone.

When the internet arrived, each component of the newspaper bundle—sports scores, financial data, local news—was split into specialized websites and apps. This unbundling eroded the newspaper’s value proposition, driving a 45% decline in weekday circulation from 122 million to 68 million and eliminating roughly 2,100 titles. The disruption mirrors earlier media shifts, showing how technology can dismantle entrenched bundles and force industries to reinvent revenue models.

Artificial intelligence now sits at a similar crossroads. Large foundation models act as all‑purpose bundles, but firms like Apple are distilling them into smaller, task‑specific models that run on devices, preserving privacy and speed. The labor impact hinges on bundle strength: weak‑bundle jobs (e.g., routine data entry) may see tasks peeled away, while strong‑bundle roles (e.g., radiologists) retain integrated responsibilities, with AI augmenting rather than replacing humans. Recognizing these dynamics enables businesses to anticipate where AI will create new value streams versus where it will merely reallocate existing work.