# AI news brief: Sunday, November 8, 2026

11 stories, most important first. Status is how solid each one is: official, confirmed, reported or rumor.

- [NVIDIA researchers show frequent context compaction can match full-history agent runs](https://ai-news.ikangai.com/story/nvidia-researchers-show-frequent-context-compaction-can-match-full-history-agent/index.md): official. NVIDIA researchers report that compacting agent context often matches full-history baselines on long tasks while cutting median peak context from 32K to 8K tokens.
- [Coinbase CEO Brian Armstrong says always-on AI agents now run his workday](https://ai-news.ikangai.com/story/coinbase-ceo-brian-armstrong-says-always-on-ai-agents-now-run-his-workday/index.md): reported. Brian Armstrong says he has reorganised his workday around a hierarchy of always-on AI agents that feed him one queue of decisions, with humans still approving every change.
- [Skill++ rewrites new skill metadata so agent routers stop displacing older skills](https://ai-news.ikangai.com/story/skill-rewrites-new-skill-metadata-so-agent-routers-stop-displacing-older-skills/index.md): official. Researchers from UIC, McGill and other institutions propose Skill++, which rewrites a new skill's metadata so agent routers keep choosing the right existing skills.
- [Alibaba Research's Recursa adapts agents to new domains without changing model weights](https://ai-news.ikangai.com/story/alibaba-research-s-recursa-adapts-agents-to-new-domains-without-changing-model-w/index.md): official. Alibaba Research introduces Recursa, which builds a validated skill repository from task experience so a model can adapt to a new domain with fixed weights.
- [Meta FAIR proposes Routed Continual Learning to split agent memory into clusters](https://ai-news.ikangai.com/story/meta-fair-proposes-routed-continual-learning-to-split-agent-memory-into-clusters/index.md): official. FAIR at Meta proposes Routed Continual Learning, which groups past agent runs into clusters with their own memory systems and cuts estimated API cost by 65% to 72%.
- [IBM's JevAct adds a runtime gate that checks agent tool calls without retraining](https://ai-news.ikangai.com/story/ibm-s-jevact-adds-a-runtime-gate-that-checks-agent-tool-calls-without-retraining/index.md): official. IBM researchers introduce JevAct, a runtime gate between an agent and its tools that checks write calls and hand-offs while leaving the model and tools unchanged.
- [Alibaba and SCUT propose SPAR, a sales agent that tests strategies as hypotheses](https://ai-news.ikangai.com/story/alibaba-and-scut-propose-spar-a-sales-agent-that-tests-strategies-as-hypotheses/index.md): official. Alibaba Group and South China University of Technology propose SPAR, a self-evolving sales agent that refines its playbook without updating model weights.
- [Researchers release RSI-Dojo environments for training agents to run post-training](https://ai-news.ikangai.com/story/researchers-release-rsi-dojo-environments-for-training-agents-to-run-post-traini/index.md): official. RSI-Dojo is a set of executable environments built from published research, and fine-tuning Qwen models on its trajectories gains 8.74 and 16.87 PostTrainBench points.
- [Stanford's Collaborative Intelligence Benchmark tests AI teamwork across 240 tasks](https://ai-news.ikangai.com/story/stanford-s-collaborative-intelligence-benchmark-tests-ai-teamwork-across-240-tas/index.md): official. Stanford researchers introduce a 240-task benchmark for teamwork skills, where the best of 18 models scores 82.2 out of 100.
- [OpenAI's LifeVal benchmark tests assistants on personal tasks built from real chats](https://ai-news.ikangai.com/story/openai-s-lifeval-benchmark-tests-assistants-on-personal-tasks-built-from-real-ch/index.md): official. OpenAI introduces LifeVal, a set of 169 agent tasks based on 148 life tasks ranked by 3,837 US consumers, where the most common failure is a wrong answer or recommendation.
- [Artificial Analysis launches Decision Index v2 with calibration-weighted scoring](https://ai-news.ikangai.com/story/artificial-analysis-launches-decision-index-v2-with-calibration-weighted-scoring/index.md): official. Artificial Analysis has released version 2 of its Decision Index, with 45% of the score now built on private datasets and a headline metric that rewards calibrated confidence.

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.
