Social Media Briefing — April 1, 2026

OpenAI announces closing its latest funding round: $122 billion in committed capital at $852B post-money valuation.

John Carmack provides a detailed technical review of the LeWorldModel paper, which applies JEPA (Joint-Embedding Predictive Architecture) to world models for robotics. He analyzes the architecture choices, SigReg loss, latent dimensions, predictor design, dropout, planning via CEM, and shares his own experience trying JEPA on Atari games.

Andrew Ng writes a lengthy post about anti-AI coalitions manipulating public sentiment. Discusses how different fear messages (extinction, warfare, environment, jobs, children) are tested for effectiveness. Supports White House federal preemption framework for AI regulation. Warns against state-level regulations that could stifle AI, drawing parallels to how anti-nuclear propaganda led to millions of pollution deaths.

SVPino reports that Claude Code's source code was leaked, criticizing the AI-writes-everything-without-review culture. Very high engagement suggests broad concern.

Research_FRI shares results of a comprehensive study on how economists and AI experts think AI will affect the US economy. Key findings: economists predict major AI progress but no dramatic economic shift. Under rapid AI progress, GDP growth hits 3.5%, labor force participation drops to 55% (~10M fewer jobs), and top 10% would hold 80% of wealth by 2050.

Karpathy reports a new npm supply chain attack on axios (300M weekly downloads), shares personal near-miss, and argues package managers need better defaults to prevent infection spread through unpinned dependencies.