With a proposal that totally reverses the existing engagement-driven approach, Marc Andreessen has ignited a new discussion about the future of social media. Andreessen proposed a social network whose feeds are prioritised according to forecast accuracy rather than incentivising indignation, virality, or constant posting.

Instead of generating clicks under this strategy, people become visible when they consistently make accurate projections. Discussions about the concept quickly spread within AI, cryptocurrency, and decentralised social circles. Projects like Trueo_app, hivy, AuspexTerminal, and Gensyn AI's Delphi entered the discourse as potential early examples of online predictive reputation systems.

• Marc Andreessen’s Idea Replaces Engagement with Accuracy
• Why Projects Like Delphi & Trueo_app Entered the Conversation?
• The Debate Around PnL & Sharpe Ratio as Reputation Metrics
• The Biggest Challenge is Turning Human Judgment Into Verifiable Signal

Marc Andreessen’s Idea Replaces Engagement with Accuracy

Social media companies have optimised feeds based on interaction numbers for almost two decades. Ranking algorithms now rely on likes, comments, shares, watch time, and repost velocity. By adding a reputation layer based on predictive correctness, Andreessen's idea completely contradicts that framework.

The main idea is basic but disruptive. The platform would reward users who consistently make correct predictions about future events rather than those who are boisterous, emotive, or controversial. Instead of influencing farming, visibility would be linked to forecasting ability.

This significantly alters social media's incentive structure. A user would be given more ranking authority inside the network if they were able to anticipate market movements, political outcomes, technical advances, sports outcomes, or macroeconomic shifts with accuracy over time. A person who consistently posts viral content without any predictive value would become irrelevant.

Communities already experimenting with decentralised intelligence systems and information markets found resonance with the notion right away. It was characterised by many observers as a potential advancement over algorithmic feeds that now give priority to emotional responses rather than signal quality.

The concept also aligns with Andreessen's long-held conviction that software systems will someday restructure entire sectors around quantifiable efficiency. In this instance, the quantifiable result is forecast accuracy.

Why Projects Like Delphi & Trueo_app Entered the Conversation?

Since some projects are already developing systems related to predictive reputation, the conversation swiftly moved beyond theory.

Delphi from Gensyn AI became one of the most frequently cited examples. Instead of merely serving as passive tools, AI models actively participate in prediction settings in Delphi's information market. Forecasts are valued by models, who are rewarded when their predictions come true. The system successfully converts intellect into an asset that can be measured economically.

In a significant aspect, that approach is similar to Andreessen's social feed idea. Instead of using popularity as a criterion, both systems try to rank entities based on their proven forecasting performance.

One sports prediction market reportedly gathered over 87,000 traders and generated $4.88 million in volume during Delphi's testnet phase, while an Oscars-related market attracted over 45,000 players. Because they demonstrated actual engagement with prediction-based coordinating systems rather than speculative interest, these individuals were crucial to the discussion.

Discussions also brought up other names, such as Trueo_app, hivy, and AuspexTerminal, as instances of new infrastructure centred on signal verification, reputation, or forecasting.

The more general consequence is that proven informational performance may eventually replace follower numbers as the primary indicator of social identity on the internet. Rather than using attention extraction to build audiences, users would utilise track records to establish a reputation.

That is an entirely separate online reward system.

Interesting. https://t.co/xKHgce8tiB— Marc Andreessen 🇺🇸 (@pmarca) May 8, 2026