Jevons Paradox

• OpenAI "Project Veritas" Launch (August 2026) — OpenAI has transitioned from a pure generative play to a "Signal-as-a-Service" model. Veritas is a suite of tools for publishers to cryptographically sign human-authored content at the point of creation. This shifts OpenAI’s positioning from a "slop generator" to a gatekeeper of authenticity, attempting to own the "Trust Layer" of the internet.
• Substack’s "Expert Graph" Protocol (Shipping Q4 2026) — Substack announced a decentralized curation protocol that allows high-signal writers to "vouch" for content across the web. By leveraging human reputation to filter AI noise, Substack is positioning itself as a high-moat alternative to Google Search, which continues to struggle with synthetic SEO spam.
• Google Search "Human-First" Indexing (Announced July 2026) — In a radical pivot, Google has begun deprioritizing pages that lack verified authorship metadata (C2PA). This is a direct response to the Jevons Paradox: as AI made SEO content cheaper, the quality of search results degraded to a breaking point, forcing Google to abandon its "crawl everything" philosophy in favor of "crawl the verified."
• Anthropic’s "Constitutional Curation" Agents (Beta Release Q3 2026) — Anthropic is testing personal AI agents that act as a "noise buffer" for users. Unlike traditional feed algorithms designed for engagement, these agents use "Constitutional AI" to filter for truthfulness and depth, potentially disintermediating the ad-supported attention economy.
• The "Dead Internet" Congressional Hearings (August 14, 2026) — U.S. lawmakers and tech executives (including Sam Altman and Dario Amodei) met to discuss the "Synthetic Tsunami." The discussion focused on mandatory "Proof of Personhood" for social media accounts to prevent AI-driven consensus manufacturing, a move that would fundamentally alter the cost structure of digital influence.

• Signal-to-Noise Inversion [HIGH] — As the cost of producing a "convincing" article or video drops to near-zero, the utility of uncurated feeds (X, TikTok, Google) is declining. Evidence: A 40% YoY increase in "Direct-to-Creator" traffic as users bypass platforms to find trusted voices.
  • Disrupted: Legacy SEO-dependent media (Dotdash Meredith, IAC), generic content mills.
  • Winners: Niche curation platforms (Substack, Beehiiv), Discord-based private communities.
  • KPI Signpost: Monitor the "Verified-to-Synthetic" ratio in top 10 Google Search results for high-intent queries (e.g., "best health insurance").
• The Rise of "Proof of Personhood" (PoP) [MEDIUM] — With AI capable of passing Turing tests in text and voice, "Identity" is the new scarcity. Evidence: Rapid adoption of Worldcoin (Tools for Humanity) and similar biometric-linked digital IDs for platform access.
  • Disrupted: Anonymous social platforms, traditional bot-detection software (which AI now bypasses).
  • Winners: Biometric identity providers, hardware-based security key manufacturers (Yubico).
  • KPI Signpost: Watch for "PoP-required" gates on major social media API access points.
• Agentic Filtering [HIGH] — The shift from "Pull" discovery (Search) to "Push" curation (Agents). Users no longer browse the web; their agents fetch and summarize verified signal.
  • Disrupted: Ad-tech incumbents (Alphabet, Meta) whose business models rely on "eyeballs on ads."
  • Winners: Agent-first platforms (Perplexity, Inflection AI, Apple Intelligence).
  • KPI Signpost: Track the percentage of "headless" web traffic (requests made by AI agents rather than browsers).

• Strengthening Moats: The New York Times / Wall Street Journal. By maintaining a closed loop of verified, human-in-the-loop reporting and proprietary data, these incumbents have turned their "legacy" high-cost structure into a premium signal moat that AI cannot replicate without permission.
• Eroding Moats: Pinterest and generic "Inspiration" platforms. These platforms are being flooded with AI-generated imagery that is visually perfect but lacks the human "taste" or "utility" users previously sought. Their algorithmic moats are being "Jevon-ized"—the abundance of images has made the platform's core value proposition commodity-grade.
• Emerging Moats: Worldcoin (Tools for Humanity). By creating a hardware-backed "Proof of Personhood," they are building a moat around the human itself. In a world of infinite AI content, the ability to prove "this was seen/liked/written by a real human" is a new, defensible infrastructure layer that did not exist in a meaningful way 12 months ago.

Counter-Thesis: The Attention Ceiling

The Jevons Paradox assumes that as a resource becomes more efficient, we use more of it. However, the Attention Ceiling argues that human time is a non-elastic resource. Unlike coal or data storage, we cannot consume 1,000x more content just because it is 1,000x cheaper to produce.

• The Argument: Incumbent platforms (Meta, Google) will successfully absorb the AI shock by using AI themselves to filter the "slop" they created. The "moat" remains the user's existing habit and the platform's massive compute advantage in running real-time "quality" classifiers. In this view, the "noise" is a temporary friction, not a structural collapse.

1. Monitor the "Curation Premium" in SaaS Pricing — Evaluate companies like Beehiiv or Patreon to see if they can command higher take-rates as users flee "free" but noisy platforms. Signal: A shift from ad-supported to subscription-heavy revenue mixes in mid-tier media.
2. Track C2PA Hardware Integration — Investigate the supply chain of Apple and Samsung for "Secure Enclave" updates that sign photos/videos at the sensor level. This is the "plumbing" of the new trust economy.
3. Assess "Agentic SEO" Exposure — For portfolio companies reliant on organic search traffic, evaluate their "Agent-Readiness." If their content is not being cited by Perplexity or Claude’s internal search, their "moat" is already gone.