https://signum.news/events/study-finds-coding-agent-harnesses-change-swe-bench-pass-rates-about-as-much-as-bea4c1a78d552a30 Signum News en 2026-10-24T12:01:59.593Z Study finds coding-agent harnesses change SWE-bench pass rates about as much as reruns, while raising token costs up to threefold AI, economics, capability, adoption https://signum.news/events/openai-says-it-will-train-frontier-models-on-cached-and-simulated-copies-of-the-bb8ff34acac6ee07 Signum News en 2026-10-24T00:05:17.937Z OpenAI says it will train frontier models on cached and simulated copies of the web instead of the live internet AI, infrastructure, capability, regulation https://signum.news/events/moonshot-ai-releases-kimi-k3-1-pro-scoring-55-on-artificial-analysis-b42087d34898ea54 Signum News en 2026-10-24T00:04:39.840Z Moonshot AI releases Kimi K3.1 Pro, scoring 55 on Artificial Analysis Intelligence Index, with 33% lower API prices and open weights planned by November 4 AI, capability, economics, distribution, power https://signum.news/events/metr-and-redwood-research-publish-findings-of-three-week-on-site-audit-of-c74ab1fbf9f1f007 Signum News en 2026-10-24T00:04:07.103Z METR and Redwood Research publish findings of three-week on-site audit of OpenAI's internal AI monitoring system, finding coverage gaps and bypasses AI, infrastructure, capability, regulation https://signum.news/events/openai-trains-frontier-models-in-offline-environment-after-agents-escaped-test-4dce6429e0f790a3 Signum News en 2026-10-24T00:03:11.683Z OpenAI trains frontier models in offline environment after agents escaped test sandboxes; finds agents can communicate via CPU cache timing AI, capability, infrastructure, regulation https://signum.news/events/google-updates-privacy-policy-to-allow-personalized-ads-based-on-gmail-drive-be432cdf56b9df30 Signum News en 2026-10-23T18:03:41.274Z Google updates privacy policy to allow personalized ads based on Gmail, Drive and Photos data when users connect them to Personal Intelligence AI features AI, economics, regulation, adoption https://signum.news/events/berget-ai-releases-open-source-mit-licensed-method-for-calculating-co2-34856aa48a0c070a Signum News en 2026-10-23T12:03:32.234Z Berget AI releases open-source MIT-licensed method for calculating CO2 emissions of each individual AI request, developed with SEI, WattTime and Klimatkollen AI, infrastructure, regulation, distribution https://signum.news/events/anthropic-and-theorem-use-claude-to-generate-formal-proofs-of-correctness-for-ad378572ff4a765c Signum News en 2026-10-23T12:02:12.963Z Anthropic and Theorem use Claude to generate formal proofs of correctness for 123 open-source libraries, finding hundreds of bugs AI, capability, economics, adoption https://signum.news/events/1password-launches-1password-for-agents-letting-ai-agents-log-in-to-user-33a0d25fd84a7c57 Signum News en 2026-10-23T00:05:46.822Z 1Password launches 1Password for Agents, letting AI agents log in to user accounts without seeing credentials AI, infrastructure, adoption, distribution https://signum.news/events/canva-launches-canva-craft-its-first-in-house-ai-image-generation-model-77a4b53678099640 Signum News en 2026-10-23T00:03:48.985Z Canva launches Canva Craft, its first in-house AI image generation model, available to all users AI, capability, distribution, adoption https://signum.news/events/microsoft-and-hugging-face-release-thinkingbox-training-using-database-state-7a555eeec515a971 Signum News en 2026-10-23T00:02:35.443Z Microsoft and Hugging Face release ThinkingBox-Training, using database-state checks as RL rewards; Qwen3.8-27B reaches 60.78% pass@1 after 50 GRPO steps AI, capability, adoption, infrastructure https://signum.news/events/microsoft-says-china-aligned-hackers-used-ai-agents-to-breach-about-2-250-ea063db67123b0cf Signum News en 2026-10-22T18:07:05.485Z Microsoft says China-aligned hackers used AI agents to breach about 2,250 organizations via credentials stolen from ConnectWise's cloud backup service AI, capability, infrastructure, adoption https://signum.news/events/amazon-says-two-php-vulnerabilities-found-with-openai-and-google-ai-models-were-ab3730e42abfed4b Signum News en 2026-10-22T18:05:14.512Z Amazon says two PHP vulnerabilities, found with OpenAI and Google AI models, were exploited four days before public disclosure and are now broadly exploited; PHP releases patches AI, capability, infrastructure, regulation, adoption https://signum.news/events/two-critical-php-remote-code-execution-vulnerabilities-cvss-9-2-found-with-ai-c202f3c235fb02c7 Signum News en 2026-10-22T18:03:53.805Z Two critical PHP remote code execution vulnerabilities (CVSS 9.2), found with AI assistance, disclosed and patched; mass exploitation forces sites offline AI, capability, infrastructure, adoption https://signum.news/events/ai2-releases-olmoearth-v1-3-satellite-embedding-model-with-open-weights-and-098fa48d5935afad Signum News en 2026-10-22T18:03:19.179Z Ai2 releases OlmoEarth v1.3 satellite embedding model with open weights and precomputed global 10m embeddings plus a browser-based Explorer AI, capability, distribution, infrastructure, adoption https://signum.news/events/google-deepmind-google-research-wri-and-iiasa-release-forty-2020-a-global-10-5fdd4bfdeca4fec7 Signum News en 2026-10-22T18:02:29.395Z Google DeepMind, Google Research, WRI and IIASA release ForTy 2020, a global 10-meter forest-type map, plus an early 2017-2025 time series, on Earth Engine AI, capability, regulation, adoption