Apple releases RSI-Inference: evaluating agents for self-improving LLM inferencetwitter.com·2 pts·low-rank·0