XNews

Dream-RSI: Self-Improving AI Loop Reduces Agent Calls 162x

Dream-RSI: Self-Improving AI Loop Reduces Agent Calls 162x

Google and DeepMind presented Dream-RSI, a recursive self-improvement loop where an AI replays past discoveries to test exploration strategies at low cost. In algorithm design, math optimization, and GPU kernel engineering, it maintained or improved quality while reducing agent invocation calls by up to 162 times.

Source posts