AgentsMastra Observational Memory: 94.87% LongMemEval SOTAMastra Blog·high signalXBlueskyLinkedInCopy linkMastra v1.2 introduces Observer+Reflector background agents compressing conversations 5-40x. 94.87% on LongMemEval with gpt-5-mini. Cuts agent memory costs 10x.SourceSource pageMastra Blog↳ Follow the threadShared entity / Stack layerMastra 1.64.0 adds pre-cloned sandbox templates and a Vitest runner so evals can gate CIGitHub ReleasesStack layer / Threat patternCROSS-CATEGORY: Four Products Shipped Pre-Execution Assurance for Agent Tools in 48 HoursTDQS, Apify, Product Hunt and StartupCorners digestStack layer / Threat patternCrewAI 1.15.19 downgrades its own telemetry from core count to a coarse machine-size bandGitHubPolicy dependency / Stack layerDRACO Redistributes One Rubric Score Across Agent Steps and Beats Ground-Truth-Reward GRPO by 5.3 Points on AppWorldarXiv 2609.04094Policy dependency / Stack layerLLMs Make More and Larger Edits on Code Written by a Different LLMarXiv 2609.03894Stack layer / Threat patternAlcaTRAz Defends Jailbreaks With Character-Level Perturbation Rules and No Model Access, Beating Llama Guard on 73.4% of CombinationsarXiv 2609.03693Policy dependency / Stack layerMoadim Ships an MIT-Licensed Scheduler That Runs Coding Agents on a Cron Instead of a PromptMoadim, via Hacker News Show HNPolicy dependency / Stack layerCopilot CLI 1.0.83 stable lets a custom agent list several models and fall through them in orderGitHub