🤝In Partnership: Meet the people building modern software

🤿Deep Dive: OpenAI expands Codex beyond coding

🤝Powered by: A morning habit people are trying

🤿Deep Dive: Anthropic launches Claude Opus 4.7

⚒Tool Snapshots: Tools for AI, no-code, and productivity

🖼AI Art: Examples of great and trending AI art

Background computer use lets Codex see, click, and type with its own cursor across any app on Mac, and multiple agents can work in parallel without getting in the way of what you are doing. This is especially useful for testing apps or iterating on frontend changes in tools that do not expose an API.

The new in-app browser lets you drop comments directly onto web pages to give the agent precise instructions, which is already useful for frontend and game development with plans to expand to full browser control beyond localhost apps over time.

Codex can now generate and refine images using gpt-image-1.5, pulling together screenshots, code, and visuals in one workflow to help with mockups, product concepts, and game assets without switching tools.

Memory is launching in preview, allowing Codex to hold onto preferences, corrections, and context it took time to gather in earlier sessions so future tasks start faster and at a higher quality without needing lengthy custom instructions each time.

Automations now support reusing existing conversation threads and self-scheduling, meaning Codex can wake itself up to continue long-running work automatically. Teams are already using this to close out open pull requests, follow up on tasks, and stay on top of activity across Slack, Gmail, and Notion.

Opus 4.7 follows instructions far more literally than previous models, which is a genuine improvement but also worth flagging since prompts written for earlier Claude models that relied on loose interpretation may now produce unexpected results. Developers are encouraged to revisit and retune their existing prompts and harnesses.

Vision gets a major boost with support for images up to 2,576 pixels on the long edge, more than three times the resolution of prior Claude models. This opens up more precise use cases like reading dense screenshots in computer use agents, extracting data from complex diagrams, and any work that needs pixel level accuracy.

On real world tasks Opus 4.7 scores at the top of the Finance Agent evaluation and on GDPval-AA, a third party benchmark covering economically valuable knowledge work across finance, legal, and other domains. Internal testing showed it producing more rigorous financial analyses, more professional presentations, and tighter cross-task integration than Opus 4.6.

File system-based memory is noticeably improved, with Opus 4.7 retaining key notes and context across long multi-session work so that follow on tasks need less setup each time. This makes it better suited for extended projects where continuity across sessions actually matters.

On cybersecurity Anthropic is using Opus 4.7 as the first model to test new safeguards aimed at detecting and blocking high risk or prohibited cyber uses. Its capabilities in this area are intentionally less advanced than Claude Mythos Preview, and what Anthropic learns from this real-world deployment will inform how it eventually rolls out Mythos-class models more broadly.

Opus 4.7 is available now across all Claude products and via the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry at the same pricing as Opus 4.6 which is $5 per million input tokens and $25 per million output tokens.

💻 Claude Code Desktop - Efficient, comprehensive coding and shipping app..

⚡ Resend CLI 2.0 - Unleashing AI power & automation from terminal.

🔄 Google Chrome Skills - Transform AI prompts into reusable workflows.

🔊 Google Gemini 3.1 Flash TTS - Multilingual, multi-speaker voice development tool.

📤 Gemini CLI Subagents - Delegate complex tasks for efficient automation.