GPT-5-Codex is now available in Cursor

AI Product Launches & Updates

• GPT-5-Codex integration: Cursor AI @cursor_ai announced that GPT-5-Codex is now available in Cursor, empowering developers with a cutting-edge coding assistant.

• Qwen3-VL open-sourced: Alibaba Qwen @Alibaba_Qwen unveiled that the flagship Qwen3-VL-235B-A22B vision-language model is now open-sourced in both Instruct and Thinking versions, outperforming peers on key vision benchmarks.

• Qwen3-Max launch: Alibaba Qwen @Alibaba_Qwen introduced Qwen3-Max with enhanced coding and agentic capabilities, featuring the Max-Instruct mode that rivals top models on SWE-Bench and Tau2-Bench.

AI Tools & Applications

• Design-to-code with Claude: Claude @claudeai demonstrated how Claude Code integrates with Figma via MCP to translate mockups into production-ready code at the component level.

• Context Engineering best practices: LangChainAI @LangChainAI announced an upcoming webinar with experts from LangChain and ManusAI, covering proven patterns and blueprints for production-grade GenAI agents.

• Composite Evaluators in LangSmith: LangChainAI @LangChainAI introduced Composite Evaluators in LangSmith, enabling weighted aggregation of multiple evaluation scores for a unified performance metric.

Product Management Insights & Strategies

• How to leverage the Product Pass: Lenny Rachitsky @lennysan shared a 2-part series detailing how he uses each tool in his Product Pass, offering actionable tips for PMs to adopt them effectively.

• Avoiding context overload: Nurijanian @nurijanian highlighted research showing that excessive context can harm AI model performance and distilled 12 actionable rules PMs can apply immediately.

• AI Evals & Discovery podcast: Teresa Torres @ttorres discussed the essentials of defining, measuring, and operationalizing AI evaluation metrics to validate product performance.

AI Industry Developments & News

• Expanding AI compute: OpenAI @OpenAI announced five new Stargate sites in partnership with Oracle and SoftBank, accelerating their 10-gigawatt infrastructure commitment ahead of schedule.

• Frontier Safety Framework updates: Demis Hassabis @demishassabis revealed enhancements to the Frontier Safety Framework, expanding risk domains and refining assessment protocols for advanced AI systems.

• Time-series foundation models: Google Research @GoogleResearch presented a new approach leveraging continued pre-training for few-shot time-series forecasting, matching supervised fine-tuning performance without extra training steps.