Social Media Briefing — May 26, 2026

Ethan Mollick argues we need hard problem repositories beyond math for AI to tackle, including engineering, economics, biology with proper evaluation criteria.

Francois Chollet argues AI should not be framed as productivity booster but as enabling new ways of working.

Marcus characterizes Musk-Altman-LeCun-Hassabis dynamics, calling Hassabis one of few decent AI figures.

Gebru argues Vatican should have demanded Anthropic stop harmful practices instead of partnering.

AlphaSignalAI summarizes a paper proposing the Implicit Curriculum Hypothesis: pretraining follows a consistent order across model families from copying to morphology to arithmetic to complex reasoning, tracked across 91 tasks and 9 models from 410M to 13B parameters.

Jerry Liu sharing a 116-page workshop deck tracing evolution of RAG, document context, and AI agents from 2023 to 2026, including pain points of naive RAG, reranking importance, agent loop offloading, document parsing challenges, and modern agent workflows

Tunguz argues Pope Leo's first encyclical Magnifica Humanitas focusing on AI is a major signal that the Catholic Church takes AI challenges seriously and hopes other institutions follow suit.

Commentary on US vs Germany GPU capacity gap, predicting US will soon have more GPUs than people, framing GPU capacity as new industrial capacity.

Hugging Face CEO Clement Delangue argues that the most important AI risk is concentration of power, capabilities, and economic gains, replying to David Sacks.