DispatchChinese Open-Source Frontier Push Qwen 3.5 GLM 5 MiniMax M2.5Interconnects·high signalXBlueskyLinkedInCopy linkThree Chinese flagship releases. Qwen 3.5 (397B MoE, Sonnet-level). GLM-5 (744B, 77.8% SWE-bench). MiniMax M2.5 RL agentic.SourceSource pageInterconnects↳ Follow the threadShared entity / Stack layerIBM Research measured agent memory as a dose, not a switch: +16.1pp for weak models, 0.0pp for GLM-5Hugging Face Blog (IBM Research)Stack layer / Threat patternHow you lay out your repo changes prompt-injection success: highly modular workspaces measurably lower attack success ratearXiv 2608.14876Stack layer / Threat patternGuillermo Rauch: Vercel Ran Cybersecurity Evals on GLM 5.3 and Calls It 'the New Open Frontier' — Cheap Enough to Scan 3× More OftenX/TwitterStack layer / Threat patternCTIFoundry: Restructuring the Corpus, Not the Model, Adds +0.19 to +0.28 F1 at Half the Tool CallsarXiv 2608.18613Policy dependency / Threat patternOpenAI Launches an Initiative on Democratic Oversight of AI in National SecurityOpenAI BlogStack layer / Update threadTrain the harness, not just the model: 6K examples of harnessed agentic RL lifts Qwen3.5-9B on SWE-bench Verified from 41.8% to 56.4%arXiv 2608.17528Stack layer / Update threadMultica Shipped Three Releases in 24 Hours and Added a MiniMax Code ACP Runtime to Its 20-Agent Issue BoardGitHubStack layer / Update threadCROSS-CATEGORY: Skills Libraries Are Becoming the Extension Point in Design and Work Tools, Replacing the Plugin MarketplaceMiniMax Design, Figma Release Notes, Asana investor release (Aug 20, Aug 13, Jun 4, 2026)