🤖 AI Agents Weekly: Claude Opus 4.7, Codex Everywhere, Claude Design, Windsurf 2.0, Qwen3.6-35B-A3B, AiScientist, and More
Claude Opus 4.7, Codex Everywhere, Claude Design, Windsurf 2.0, Qwen3.6-35B-A3B, AiScientist, and More
In today’s issue:
Anthropic ships Claude Opus 4.7
Codex extends to Mac apps
Claude Design enters research preview
Windsurf 2.0 delegates to Devin
Qwen drops 3.6-35B-A3B open weights
OpenAI Agents SDK adds sandboxes
Gemini CLI adds subagents
FrontierSWE benchmark launches
NVIDIA releases Nemotron 3 Super
AiScientist lifts long-horizon research
And all the top AI dev news, papers, and tools.
Top Stories
Claude Opus 4.7
Anthropic released Claude Opus 4.7, its most capable Opus model yet, built for long-running agentic work with more rigorous self-verification and tighter instruction following. Opus 4.7 also powers the new Claude Design product and Anthropic’s Glasswing cybersecurity frontier model.
Self-verifying long-running work: Opus 4.7 checks its own outputs before reporting back and handles multi-hour tasks with less supervision, making it a stronger default for hand-offs where the agent must own the full loop.
Vision upgrade: The model sees images at more than three times the resolution of Opus 4.6 and produces higher-quality interfaces, slides, and documents, which is the foundation for the new Claude Design research preview.
New reasoning and budget controls: A new xhigh effort level between high and max gives developers finer latency/quality tradeoffs on hard problems. Task budgets (beta) let Claude prioritize work and manage cost across longer runs.
Claude Code upgrades: A new /ultrareview command runs a dedicated review pass over changes that flags what a careful reviewer would catch, and auto mode is now extended to Max users so long tasks run with fewer interruptions.