This paper presents Dream-RSI, a framework that lets AI agents improve their exploration strategies by learning from past discovery attempts without constantly running expensive new experiments. The system uses historical search data as a simulator to test and refine exploration policies cheaply before deploying them back into the real search space.

• Solves the exploration bottleneck in recursive self-improvement by creating a replay simulator from accumulated discovery history, enabling off-policy feedback without costly online re-evaluation
• Keeps the exploration layer separate from the underlying agent, making strategies explicit and programmable while maintaining compatibility with existing systems
• Demonstrates competitive or better results across algorithm engineering, mathematical optimization, and GPU kernel engineering while substantially cutting discovery costs

In Robot-Proof, neuroscientist Vivienne Ming argues AI will “deprofessionalize” law, medicine and finance, reducing high-skill roles to assembly-line tasks. She calls for a shift from knowledge transmission to building “meta-learning” traits—cognitive flexibility, empathy, resilience and divergent thinking—through project-based education and capacity-driven hiring. Drawing on proprietary HR data, she proposes US-centric reforms and data trusts but offers little on collaborating with frontline educators and clinicians.

• AI will "deprofessionalize" law, medicine and finance by breaking high-skill work into assembly-line tasks, not just automating low-skill jobs
• Ming argues education and hiring should shift from knowledge transmission to measuring capacities like resilience and creative problem-solving via game-style assessments, citing pilots in finance hiring and adaptive schooling
• Critique: her proposals lean on proprietary HR data and US-centric reforms, with little attention to input from frontline educators and clinicians