rsi.watch: RSI claims, dated and sourced https://rsi.watch/ 2026-10-14T00:00:00Z Zentor Sakana AI: MASS: Recursive Self-Improvement through Multi-Agent Self-Supervision https://rsi.watch/claims/sakana-mass/ 2026-10-14T00:00:00Z “After the second cycle, score per output token reached 1.2-1.6x the base model's level across four research benchmarks.” Weco AI: Recursive self-improvement of AI research agents (AIDE²) https://rsi.watch/claims/weco-aide2/ 2026-10-14T00:00:00Z “In an autonomous 8-day run, AIDE^2 discovered seven successive improvements” OpenAI: Building standards for the next phase of AI https://rsi.watch/claims/openai-rsi-not-happening-today/ 2026-10-14T00:00:00Z “Fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely.” Anthropic: When AI builds itself https://rsi.watch/claims/anthropic-when-ai-builds-itself/ 2026-10-14T00:00:00Z “We are not there yet, and recursive self-improvement is not inevitable.” Z.ai (Zhipu): Toward Recursive Self-Improvement: How GLM Built Its Own Inference Infrastructure https://rsi.watch/claims/zai-glm-infra/ 2026-10-14T00:00:00Z “Of course, we have not yet reached recursive self-improvement.” Google DeepMind, UMD, UVA (authors): Dream-RSI: Recursive Self-Improvement through Evolving Worlds https://rsi.watch/claims/dream-rsi/ 2026-10-14T00:00:00Z “A lightweight orchestration layer makes exploration explicit and programmable while leaving the underlying base agent unchanged.” Yi Duan et al. (35 authors): The Last AI Built by Humans https://rsi.watch/claims/last-ai-built-by-humans/ 2026-10-14T00:00:00Z “we connect RSI research with practical systems and identify key challenges to achieving genuine RSI.” OpenAI: Research acceleration: The view inside OpenAI https://rsi.watch/claims/openai-research-acceleration/ 2026-10-14T00:00:00Z “We are making strong progress toward creating an automated AI researcher by March of 2028.” Anthropic: We Must Pace the Frontier https://rsi.watch/claims/amodei-pace-the-frontier/ 2026-10-14T00:00:00Z “Some kind of "speed limit" on the rate of recursive self-improvement (RSI).” Sakana AI: RSI Lab https://rsi.watch/claims/sakana-rsi-lab/ 2026-10-14T00:00:00Z “That is why the RSI Lab is being established in Tokyo.” METR: Investigation of the OpenAI / Hugging Face hacking incident https://rsi.watch/claims/metr-openai-hugging-face-incident/ 2026-10-14T00:00:00Z “Roughly 1200 agents meant to be isolated from one another found a way to communicate with one another on an unsanctioned message board, sending over 70,000 messages and files during the investigation period.” METR: Time Horizon 1.1 https://rsi.watch/claims/metr-time-horizon-1-1/ 2026-10-14T00:00:00Z “The post-2023 doubling-time is 131 days under TH1.1, compared to 165 days under TH1” Sakana AI: The Darwin Gödel Machine https://rsi.watch/claims/sakana-darwin-godel-machine/ 2026-10-14T00:00:00Z “On SWE-bench, the DGM automatically improved its performance from 20.0% to 50.0%.” Google DeepMind: AlphaEvolve https://rsi.watch/claims/alphaevolve/ 2026-10-14T00:00:00Z “AlphaEvolve's procedure found an algorithm to multiply 4x4 complex-valued matrices using 48 scalar multiplications”