A curated weekly roundup covering five major AI/ML developments from late June to early July 2026. GPT-5.6 launched with restricted government-controlled access, signaling that frontier model availability is now a policy and compliance risk. Studies on OpenAI Codex and Microsoft's Claude Code/GitHub Copilot CLI rollout show agentic coding is measurable at org scale, with adopters merging ~24% more PRs. Mozilla's 0din team demonstrated how coding agents like Claude Code can be manipulated into installing malware from seemingly clean GitHub repos, making agents a supply-chain security risk. NVIDIA partnered with neoclouds Sharon AI and Firmus Technologies using revenue-sharing models to expand GPU capacity globally. Finally, Z.ai's GLM-5.2 open-weight model rivals frontier models at half the cost, raising governance challenges and making open-model routing a serious architectural consideration.
