OpenAI Updates Codex into General-Purpose Mac AI Operating System, Launches GPT-Rosalind for Drug Discovery
Codex now controls any Mac app by seeing your screen, plus GPT-Rosalind targets drug discovery.
OpenAI has reimagined its Codex desktop application, evolving it from a coding-focused agent into a full-fledged Mac AI operating system. The upgraded Codex can now see the user's entire screen and control any desktop application—from email clients to design tools—by visually interpreting UI elements and executing multi-step tasks autonomously. This marks a major leap from its prior coding-only scope, enabling professionals to automate complex workflows without API integrations. The system uses advanced computer vision and agentic decision-making to navigate interfaces, click buttons, and input text across disparate apps, effectively functioning as an OS-level assistant. Early testers report up to 60% time savings on repetitive cross-app tasks like data extraction, report generation, and software testing.
In parallel, OpenAI announced GPT-Rosalind, a specialized AI model fine-tuned for drug discovery and molecular biology. Named after Rosalind Franklin, the model excels at predicting molecular properties, simulating protein-ligand interactions, and identifying promising drug candidates from massive chemical libraries. It leverages a custom training dataset of over 100 million molecular structures and published research, achieving 35% higher accuracy in binding affinity predictions compared to generic LLMs. GPT-Rosalind can also generate novel molecules with desired properties and suggest synthesis pathways, potentially reducing early-stage drug development timelines by 40%. The model is available via OpenAI's API for pharmaceutical companies and research institutions, with a focus on accelerating the discovery of treatments for rare diseases and antibiotic-resistant pathogens.
- Codex now controls any Mac app by visually interpreting the screen, enabling automated cross-app workflows.
- GPT-Rosalind achieves 35% higher accuracy in binding affinity predictions for drug discovery.
- The new drug model can generate novel molecules and synthesis pathways, cutting early-stage development timelines by 40%.
Why It Matters
Codex turns Mac into an agentic OS, while GPT-Rosalind could slash drug discovery timelines by years.