A curated AI research newsletter covering four major topics: (1) A large-scale Oxford/Stanford/LSE study showing AI systems are reliably more persuasive than expert humans across 18,978 conversations, with AI nearly 3x more effective than professional fundraising canvassers; (2) A discussion on timelines to self-sustaining AI that requires no human labor, with forecasters split between 10 and 50 years depending heavily on humanoid robot progress; (3) A Google DeepMind paper exploring pathways from AGI to ASI via scaling, algorithmic breakthroughs, recursive self-improvement, and multi-agent coordination; (4) Startup Recursive demonstrating early automated AI research results achieving state-of-the-art on language model training and GPU kernel optimization benchmarks. The issue closes with a speculative short story about AI-human negotiations.
